KR100287905B1 - Real time voice playback system and voice playback method using the same - Google Patents

Abstract

The present invention relates to a real-time voice reproducing system and a voice reproducing method using the same. The present invention relates to a real-time voice reproduction system and a voice reproduction method using the same. Such a real time voice reproducing system includes a voice processor comprising a voice recognition module and a voice synthesis module, a voice reproducing unit capable of reproducing voice data, a memory unit used in the voice recognition module, a voice synthesizing module and a voice reproducing unit, And a control unit for controlling the voice processing unit, the memory unit, and the voice reproducing unit.

Description

Real time voice playback system and voice playback method using the same

The present invention relates to a real-time voice playback system and a method for reproducing a voice using the same, and more particularly, to speech recognition in a speech processing board in which speech recognition, text to speech (TTS), and speech play are integrated. The present invention relates to a real-time voice reproducing system and a voice reproducing method using the same, which are suitable for real time processing of voice reproducing and voice synthesis processing in a voice guidance service.

Voice recognition boards and voice synthesis boards that are configured for voice recognition and voice synthesis (TTS) applied to additional communication systems, such as the Audio Response Service (hereinafter referred to as ARS) system And voice playback boards, respectively, have been used to process voice recognition and voice synthesis services.

Each dedicated board with a built-in processor loaded with a voice recognition board and a program loaded for the role of a voice synthesis board and a voice playback board is controlled by a CPU (central processing unit) board that controls each board. It has a configuration that runs each service.

In this way, even if one channel of voice recognition and voice synthesis service is required, each board in charge of individual service is required, and each board must be controlled, which leads to waste of system resources.

Also, in order to develop each service in real time, a high speed data bus was required for data and message communication between boards.

A voice processing system for providing such a conventional voice recognition guide service will be described with reference to the accompanying drawings.

1 is a block diagram illustrating a conventional speech processing system.

In the conventional speech processing system, as shown in FIG. 1, each dedicated board is used to provide a voice recognition guide service. A voice recognition board 2 for voice recognition and a voice for inputting arbitrary sentences and synthesize a voice CPU board (1) for controlling the synthesis dedicated board (3), the speech reproduction exclusive board (4) and the speech recognition exclusive board (2), the speech synthesis dedicated board (3), and the speech reproduction dedicated board (4) for reproducing voice. It is composed of At this time, the CPU board 1, the speech recognition dedicated board 2, the speech synthesis dedicated board 3, and the speech reproduction dedicated board 4 are constituted in an audio response system 10.

The voice answering system 10 may further include a trunk card for matching with a public switched telephone network (PSTN), a personal communication system (PCS) network, and a digital communication system (DCN) network. 7) is further provided.

In this case, a method of reproducing pre-recorded data by an announcer or the like with respect to the recognized result recognized by the voice recognition board 2 has been used, and the pre-recorded data is stored in a system hard disk (HDD). The playback of the voice is made on the audio playback board (4) Serviced to users.

In this case, reference numeral 6 denotes a control signal and a message transmission bus.

In the conventional voice processing system for providing a voice recognition guide service, a call for a service is received and a voice recognition board 2 receives a voice as a digital signal from a trunk card 7 connected to the outside to a voice recognition algorithm. After the voice recognition, the result is transferred to the CPU board 1 in which the service scenario is programmed.

The CPU board 1 proceeds to the guide scenario according to the result value. The pre-recorded voice file is transferred to the voice reproduction board 4 through the data bus 5 to receive the transmitted data. One voice playback dedicated board 4 is to play in real time.

In case of playing the voice using the function of voice synthesis, the voice synthesized file is played in advance, or only after the synthesis of the inputted text is finished, even if it is required to process in real time, it is played on the voice play board 3 only. It is supposed to be.

Such a conventional voice processing system has the following problems.

First, in order to add new voices, the voice data files recorded by the announcer's voices must be stored on the system's hard disk. That is, because the speech synthesis function is not applied, there was a side in which a separate cost was paid when adding a new voice.

Second, even when the voice synthesis service is applied, it is not a real-time or because a voice synthesis board must be used, it is a waste in terms of cost and system resources.

An object of the present invention has been made in view of the above-mentioned problems of the prior art, by configuring the board so that voice recognition, speech synthesis and voice playback can be performed in parallel at the same time, voice playback and voice synthesis in the voice guidance service The present invention provides a real-time voice reproducing system capable of providing a voice recognition service and a voice synthesis service by time real-time processing, and a voice reproducing method using the same.

According to an aspect of the present invention for achieving the above object, a voice processing unit consisting of a voice recognition module and a voice synthesis module, a voice playback unit for reproducing voice data, the voice recognition module, voice synthesis module and voice And a control unit for controlling the voice processing unit, the memory unit, and the voice reproducing unit.

Preferably, the memory unit is a recognition memory unit for recognition parameters required for recognition and other memory areas for the execution of the recognition program, a database for speech synthesis, and a TTS memory for other memory areas for the execution of the TTS program. And a shared memory unit for real-time speech synthesis and reproduction in the speech synthesis unit and the speech reproduction module.

Preferably, the speech processing unit is composed of a processor loaded with a speech recognition and speech synthesis program.

Preferably, the shared memory includes a first composite data output region and a second composite data output region, so that the voice reproducing unit reads the first composite data output region and the second composite data output region separately. Play it back.

Preferably, the voice processing system may be connected to any communication channel to transmit and receive voice, to function as a call, and to receive a scenario message and text. It is further provided.

According to another aspect of the present invention for achieving the above object, preparing a voice recognition, determining whether the input voice requires the reproduction of the pre-recorded voice, the determination result of the pre-recorded voice Reproducing the pre-recorded voice file if the reproduction is requested, preparing a voice synthesized text to be synthesized and serviced if the regeneration is not required to reproduce the pre-recorded voice, and synthesizing the text to be serviced. Copying the speech synthesized text to one memory area; determining whether there is more text to synthesize; if there is more text to synthesize, playing the speech synthesized text of the one memory area and simultaneously randomly Synthesizing the remaining text in another memory area of the memory area; At the same time as the reproduction property ends comprises the step of reproducing the synthesized speech of text in the other memory area.

According to the present invention as described above, in the voice recognition ARS system, an arbitrary voice guidance service according to the recognition result may be provided through a pre-recorded voice, or may be selectively serviced in real time using a TTS.

1 is a block diagram showing a conventional speech processing system

2 is a block diagram showing a real-time voice playback system of the present invention

3 is a flowchart showing a voice reproducing method using the present invention real-time voice reproducing system

Explanation of symbols for main parts of the drawings

20: CPU board 30: dedicated voice processing board

31: external interface unit 32: voice processing unit

33 control unit 34 recognition memory unit

35: TTS memory unit 36: shared memory unit

37: voice playback unit 38: data bus

39: control signal and message bus 40: subscriber board

Hereinafter, a configuration and an operation according to an exemplary embodiment of the present invention will be described with reference to the accompanying drawings.

2 is a block diagram showing a real-time voice playback system of the present invention.

The present invention real-time voice playback system is configured in an audio response system (Audio Response System) 40, in detail, the voice processing unit 32 composed of a voice recognition module 32a and a voice synthesis module 32b, and A recognition memory unit 34 for use in the voice recognition module 32a, a TTS memory unit 35 for use in the voice synthesis module 32b, a voice reproducing unit 37 for reproducing digital voice data, and the voice Shared memory unit 36 to be shared by playback unit 37 and speech synthesis module 32b, voice processor 32, recognition memory unit 34, TTS memory unit 35, shared memory unit 36 And a control unit 33 for controlling the audio reproducing unit 37.

At this time, the speech processing unit 32 and the speech reproducing unit 37, which are composed of the speech recognition module 32a and the speech synthesis module 32b, can perform speech processing so that speech recognition, speech synthesis, and speech reproduction can be simultaneously performed in parallel. All of the dedicated board 30 is configured, and the internal memory configuration of the voice processing board 30 is logically divided into a recognition memory unit 34, a TTS memory unit 35, and a shared memory unit 36.

The voice processing unit 32 is composed of a processor in which the voice recognition and voice synthesis programs are loaded, so that the voice processing module (not shown), the voice recognition module 32a, and the voice can perform voice recognition and voice synthesis according to the message. Synthesis module 32b is programmed so that two services on a board can be selectively selected according to the message.

The recognition memory unit 34 is configured for recognition parameters necessary for recognition and other memory areas for performing the recognition program.

The TTS memory unit 35 is configured for a database for speech synthesis and other memory areas for the execution of the TTS program.

The shared memory unit 36 includes a first composite data output region 36a and a second composite data output region 36b for real-time speech synthesis and reproduction.

The two regions 36a and 36b of the shared memory unit 36 divided as described above are read out by the audio reproducing unit 37 so as to enable voice reproduction.

In addition, the external interface unit 31 may be connected to a communication channel to transmit and receive voices, to call, and to receive a scenario message and text from the CPU board 20. Receive (Text)

In addition, the voice reproducing section 37 is a module capable of reproducing voice digital data.

Reference numeral 39 is a data bus and 38 is a control and message transmission bus.

3 is a flowchart for explaining a voice processing method using the voice processing system of the present invention.

In the voice processing method using the voice processing system of the present invention, when a call is received from the ARS device and the voice processing service is requested from the CPU of the ARS to the voice processing board 30, the voice processing board 30 is connected to an external interface 31. Recognizing the incoming voice received through) to guide the service contents to the user via voice playback via the external interface 31, the following cases are implemented to be selectively made by an external message.

In the first case, the recorded voice data is read from the system's hard disk (not shown) and processed through the voice playback of the voice processing board 30 according to the scenario of controlling the service. In this case, the voice is reproduced on a dedicated voice processing board by receiving a text to be guided by a scenario of controlling a service and synthesizing the voice by using TTS.

The voice processor 32 of the voice signal processing board receives a voice through an external interface 31 to prepare voice recognition for a voice recognition service for the received call.

Subsequently, if the voice is recognized, it is determined whether it is a user voice input.

If the determination result 101 is not a voice input, the process returns to the voice recognition preparation 100 state.

However, if the determination result 101 is a voice input, the voice recognition and recognition result is reported.

At this time, the speech recognition is performed by the speech recognition module 32a of the speech processing unit 32, and the recognition result is reported to the CPU of the CPU board 33.

Subsequently, it is determined whether or not reproduction of the pre-recorded data is necessary as a result of the recognition.

In the case of the first method (when the previously recorded voice data is read from the hard disk of the system and processed through the voice playback of the voice processing board 30), the judgment result 103 is the playback of the pre-recorded voice data. Reads the voice data to be reproduced from the system hard disk by the CPU through the external interface 31 of the board 30 and stores the data in the first synthesized output data area (hereinafter abbreviated as first area) 36a of the board 30. Write

At this time, the first area 36a and the second synthesized output data area (hereinafter abbreviated as second area) 36b have a fixed size as necessary for one-time audio reproduction, so that the CPU writes the first data. Write to the first area as long as the area.

When the voice data to be played back is written to the first area, the voice player 37 starts voice playback.

At the same time as the voice is reproduced, the CPU writes the remaining voice data to be reproduced in the second area as much as the second area. After the voice is reproduced in the first area, the CPU reproduces the data of the second area.

Likewise, in the first region, the remaining audio data is stored when the second region is in the playback operation, and the above-described process is repeated until all the data is reproduced.

On the other hand, in the case of the second method (in case of playing the voice of the voice processing board 30 by synthesizing the voice by receiving the text to be guided service), the text to be serviced instead of the voice data is provided to the external interface 31. After inputting text with a certain size through text-to-speech, the data is written in the first and second areas alternately, so that cross play can be performed as in the first method.

Here, what is to be implemented in the speech synthesis module 32b for cross reproduction is not to proceed the above process after the synthesis is completed for the input text, but after the speech synthesis for the first region (or the second region) once The process of writing data in the first area (or the second area), synthesizing data of the second area (or the first area) in parallel again at the time of reproducing the data of the second area, and performing data writing You should be able to do it repeatedly.

This process is explained in detail as follows.

When the CPU requests text to be played back with the TTS service, the external interface 31 of the board 30 receives input of the requested message and text.

At this time, the input text input to the board 30 is copied to a specific area of the memory (105-1), and the text is read by the speech synthesis module 32b of the speech processor 32 to synthesize the speech. (105-2)

If it is stored beyond the amount of text that can be read by the speech synthesis module 32b, the text is first read by the specified amount and synthesized (105-2), and after the synthesis is completed, the next remaining text is read and processed. (107)

Subsequently, it is determined whether the text to be synthesized remains.

If there is no text to be synthesized as a result of the determination, the voice reproducing unit 37 reproduces the voice of the first region.

However, if the text to be synthesized remains in the determination result 108, the following operation proceeds.

That is, the speech synthesis module 32b synthesizes according to the speech synthesis algorithm. When the program structure is synthesized by the first region (or the second region), the synthesized data is converted into the first region (or the second region). (110)

In other words, when the speech synthesis starts and the synthesis data is written to the first region, the reproduction of the first region begins, and the reproduction of the second region begins and the speech synthesis module 32b synthesizes the next speech synthesis. If the synthesis continues as much as the second area, data is written to the second area.

This is possible because the reproduction time of the first region is longer than the time required for speech synthesis and data storage.

Next, it is determined whether the text to be synthesized remains.

As a result of determination 111, if there is no text to be synthesized, the text of the second area is reproduced by voice.

That is, after the reproduction of the first region is finished, the reproduction of the stored data of the second region is started.

However, if there is remaining text to be synthesized as a result of the determination 111, the voice synthesis module continues the voice synthesis again as described above and writes the remaining voice data in the first area. It continues until it is completed. (113)

As described above, the present invention can perform voice playback and voice synthesis in real time during voice guidance service (ARS) based on voice recognition results in a voice processing board integrated with voice recognition, voice synthesis, and voice playback. The voice recognition service and the speech synthesis service can be performed in real time.

Claims (8)

A speech processing unit comprising a speech recognition module and a speech synthesis module; A voice reproducing unit capable of reproducing voice data; A memory unit used in the voice recognition module, the voice synthesis module and the voice reproducing unit; And a control unit for controlling the voice processing unit, the memory unit, and the voice reproducing unit. 2. The TTS memory of claim 1, wherein the memory unit is a recognition memory unit for recognition parameters required for recognition and other memory areas for performing a recognition program, and a TTS memory for other memory areas for performing a database and a TTS program for speech synthesis. And a shared memory unit for real-time speech synthesis and reproduction in the speech synthesis unit and the speech reproduction module. The system of claim 1, wherein the speech processor comprises a processor in which a speech recognition and speech synthesis program is loaded. 2. The shared memory of claim 1, wherein the shared memory includes a first composite data output region and a second composite data output region, so that the voice reproducing unit reads the first composite data output region and the second composite data output region separately. Real-time voice playback system, characterized in that for playing the voice. The real-time voice reproducing system of claim 1, wherein the voice processing system further comprises an external interface connected to an arbitrary communication channel to receive and transmit voice, and to receive a call function and a scenario message and text. Preparing speech recognition; Determining whether the input voice requires reproduction of the pre-recorded voice; Reproducing the pre-recorded voice file if it is determined that the pre-recorded voice is to be played back; Preparing a speech synthesized text to be synthesized and serviced if the result of the determination does not require reproduction of the pre-recorded voice; Voice synthesizing the text to be serviced; Copying the speech synthesized text into one memory area; Determining whether there is more text to be synthesized; Reproducing the speech synthesized text of the one memory area while simultaneously synthesizing the remaining text in any other memory area when there is more text to be synthesized; And reproducing the synthesized text of the other memory area at the same time that the voice reproducing of the one memory area is ended. 7. The method of claim 6, further comprising: synthesizing the remaining text to be synthesized and copying the remaining text to the one memory area when there is more text to be synthesized when voice playing the synthesized text in the other memory area; And reproducing the voice of the text copied to the one memory area at the same time that the synthesized text voice reproduction of the memory area is terminated. 7. The method of claim 6, wherein, when the prerecorded voice file is larger than the one memory area, the prerecorded voice file is written to the other memory area when the prerecorded voice file is reproduced. And repeating the step of reproducing the voice file in the other memory area at the same time as the end until the pre-recorded voice reproduction is completed.