KR0153642B1 - Character-voice transformation service apparatus and control method of the same - Google Patents

Abstract

The present invention relates to an apparatus for controlling a text-to-speech service and a method of controlling the same, which are connected to information providers and converted to a voice menu and information to an electronic telephone subscriber to a voice synthesized through a telephone. Connection between the subscriber connection means 10 connected to the telephone network connection part 1, the central processing means 20 connected to the subscriber connection means 10, the central processing means 20 and the external packet network connection part 3; In the text-to-speech service apparatus and a method for controlling the text-to-speech service applied thereto, the central processing means is generated when the subscriber access means 10 generates a burner call. A first step (20) of connecting to an information provider through the packet network connection unit (3) to receive data; A second step of analyzing the received data to determine whether a current state is a step of selecting a menu unless the service is released; A third step of receiving menu data according to whether the menu selection step is performed, generating synthesized sound or detecting data information to generate and transmit a corresponding synthesized sound; And a fourth step of receiving the DTMF signal pressed by the user and transmitting a command to the information provider to move to the next state through the data processing means 30. It is effective to provide information in the form of a sound note to the user by converting to voice.

Description

Text-to-speech service device and control method thereof

1 is an overall configuration diagram of a communication network system to which the present invention is applied.

2 is a hardware configuration diagram of a text-to-speech service apparatus according to the present invention.

3 is a general software diagram of a text-to-speech service apparatus according to the present invention.

4 is a structural diagram of a speech synthesis parameter according to the present invention.

5 is a flowchart of speech synthesis processing according to the present invention.

6 is a general control flow diagram of a text-to-speech service according to the present invention.

* Explanation of symbols for the main parts of the drawings

1: Telephone network connection block 2: Text-to-speech block

3: packet network access block 10: subscriber access unit

20: central processing unit 30: data processing unit

40: speech synthesis unit

The present invention relates to an apparatus for controlling a text-to-speech service and a method of controlling the same, which are connected to information providers such as a hightel, and converted into a voice menu and information to a subscriber of an electronic telephone to a voice synthesized through telephone rules. .

The voice service currently in use has a method of recording the information to be provided by voice and storing it digitally, and then converting the information desired by the user to analog again and playing. This method has a problem that it is difficult for anyone to provide the desired information anywhere, because the information providing system can input information only if there is a recording equipment, analog / digital conversion equipment.

The present invention has been made to solve the problems of the prior art as described above, the information provided to the information providing system through a computer or a terminal used by many people to be able to provide information if anyone can input the character It is an object of the present invention to provide a text-to-speech service apparatus and a method of controlling the same, which converts the text into speech and provides the user with information in the form of speech.

In order to achieve the above object, a text-to-speech service device of the present invention is a communication network comprising a telephone network connection unit connected to a public telephone network accommodating a general telephone subscriber and a packet network connection unit connected to a public data network accommodating an information provider. In the text-to-speech service device applied to the system, the ring-signal and hook-on signals are sensed and the hook-on signal is sent to the telephone network connection to request a call disconnection, and the synthesized analog voice is converted into a pulse code modulation (PCM) signal. Subscriber access means for converting and transmitting to the telephone network connection unit; A central processing unit configured to operate a high-speed device and a multitasking OS to perform a voice synthesis function for providing a synthesized voice to the subscriber access unit in real time by controlling the overall system operation; Data processing means connected to the central processing means and the packet network connection part and performing data input / output with the outside; And a voice synthesizing means for converting a text into a voice and synthesizing the digital voice under the control of the central processing means and transmitting the synthesized voice to the subscriber access means.

In addition, the text-to-speech service control method, in the text-to-speech service control method applied to the text-to-speech service apparatus, when a call is generated from the subscriber access means, the central processing means is an information provider through the packet network connection unit; A first step of connecting with and receiving data; A second step of analyzing the received data to determine whether a current state is a step of selecting a menu unless the service is released; A third step of receiving menu data according to whether the menu selection step is performed, generating synthesized sounds or detecting data information to generate and transmit a corresponding synthesized sound; And a fourth step of receiving the DTMF signal pressed by the user and transmitting a command to the information provider to move to the next state through the data processing means.

That is, according to the present invention having the above-described configuration, a large number of information can be easily and quickly serviced, and the information enhancement database is composed of text file type characters rather than voice data, so the size of the storage medium used is also reduced. There is this.

On the other hand, in the service area of the present invention, it is being tested to make a text-to-speech connection by connecting with a hightel among various information providers in Korea. Heytel is the most common information and communication service connected through the general telephone network. Users can get and share a lot of information such as news, weather, stock information, culture / life, club meetings and so on. Of course, the service area of the device is not limited to only hightel, and anyone who is a value-added network operator can service it.

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

1 is an overall configuration diagram of a communication network system to which the present invention is applied and includes a telephone network connection block 1 connected to a public telephone network accommodating a general telephone subscriber and a text-to-speech conversion of the present invention connected to the telephone network connection block 1. A block 2, a packet network connection block 3 connected between the text-to-speech block 2 and a public data network for accommodating the information provider.

As shown in the figure, the telephone network connection block 1 is a part in charge of the telephone interface of users having a general electronic telephone. The telephone network access block 1 performs a connection with an electronic switch and sends a channel every time a call is received. It has a function of connecting with the conversion block (2).

The text-to-speech conversion block 2 is a part to which the present invention is applied, a function of connecting with the telephone network connection block 1 to send and receive connection signals, a function of driving a voice synthesis algorithm to convert text into voice, and a packet network connection. It consists of the function of exchanging data with the information provider through the block (3).

The packet network access block (3) connects the text-to-speech block (2) and the information provider while performing the X.25 setup, connection, and disconnection that connect the information database through the public data network.

2 is a hardware configuration diagram of a text-to-speech service apparatus according to the present invention, which is divided into a subscriber connection unit 10, a central processing unit 20, a data processing unit 30, and a voice synthesis unit 40.

The subscriber access unit 10 is composed of a codec for PCM (Pulse Code Modulation) signal conversion, a channel selector for setting an access channel, and a DTMF converter that detects a dual tone multifrequency (DTME) signal and converts the signal into 8 bits. Same as

First, it detects ring signal and hook-on signal and sends hook-on signal to telephone network access block (1) to request call disconnection. Second, converts synthesized analog voice into PCM (2.048MHz) signal and telephone network access block (1). ), There is transmission function, third, dial-up channel setting function, fourth, DTMF signal conversion function.

The central processing unit 20 is composed of a central processing unit (CPU), a floating point processing unit (FPU), memory devices (ROM, SRAM, DRAM, DPRAM), and a device controller, and monitors the entire hardware. This is the part where system program and application program operate. In particular, high-speed devices and multitasking operating systems (OS) can be operated to perform voice synthesis.

The configuration will be described in detail with the MC68030 CPU (33MHz), MC68882, FPU (33MHz), 1Mbyte ROM, 2Mbyte SRAM, 16Mbyte DRAM, 4K word DPRAM, and device controller.

The ROM has a system control function, the SRAM has a voice synthesis program that requires fast execution, the DRAM has a synthesis database that requires a large amount of memory, and the DPRAM exchanges data with the voice synthesizer 40. Used for

The data processing unit 30 is composed of serial input / output and Ethernet input / output, and is responsible for data input / output of the central processing unit 20 and externalization. Serial I / O implements three input / output ports using MC68901 MFP and Z8530 SCC and is used to exchange data with packet network access block (3). That is, the DTMF signal pressed by the subscriber is converted into 8-bit data and transmitted to the packet access block 3 through serial I / O, and the data transmitted from the information provider is again connected to the data processing unit connected to the packet network access block 3 ( 30 is transmitted to the central processing unit 20.

Ethernet I / O configured with AM7990 LANCE is used for debugging the system and downloading synthetic databases and applications with a capacity of about 14M bytes.

The voice synthesizer 40 converts a text into a voice, and is composed of a TMS320C30 DSP (33 MHz), a 2M byte SRAM, a 16K byte ROM, a TLS320C46 AIC, and a device controller. The speech synthesis algorithm is implemented in ROM and SRAM and handles signal processing synthesis using the synthesis parameters stored in the DPRAM from the central processing unit 20. The synthesized digital voice is transmitted to the codec of the subscriber access unit 10 through an AIC having a resolution of 14 bits every 0.0625 msec.

3 is an overall software configuration diagram of the text-to-speech service apparatus according to the present invention, which includes a TMS320C30 DSP (Digital Signal Processor) unit 101 for performing PSOLA synthesis and a VRTX32 OS unit operating in the central processing unit 20 ( 102).

The VRTX32 OS unit 102 operating in the central processing unit 20 may include a DPRAM connection driver 110, a serial connection driver 120, an Ethernet connection driver 130, a telephone connection driver 140, an interrupt service routine 150, Task management 160, system control 170, and synthetic parameter generation 180.

The DPRAM connection driver 110 is a part that allows the DSP unit 101 to transmit and receive a synthesis parameter through DPRAM, and the OxFFE address is the central processing unit using OxFFF and OxFFE addresses, which are the highest addresses in the DPRAM region of 4K words. When 20) interrupts the voice synthesis unit 40 and transmits the necessary parameters for synthesis, the OxFFF addressed by the voice synthesis unit 40 interrupts the central processing unit 20 to pass the results generated in the middle of the synthesis process. use. The rest of the DPRAM area is used to store the compositing parameters.

The serial I / O connection driver 120 manages the data transmitted and received from the serial port. Since the data transmission is less expensive, the data transmission is handled in a polling manner. It is handled by interrupt method and configured to operate at 9600bps.

The Ethernet connection driver 130 is used for debugging a system and downloading a synthetic database and an application program, which occupies a large capacity of about 14M bytes, from the workstation in the memory of the central processing unit 20, and the transmission speed is 10Mbps.

The telephone connection connection driver 140 processes hook on / off signals, ring signals, and DTMF signals coming from the subscriber connection unit 10, and reads a certain address when an interrupt is received, and then selects one of three signals. It is configured to judge and process.

The interrupt service routine 150 manages a timer interrupt for controlling the connection drivers 110-140 and the system, and informs a corresponding application program using a mailbox or a counter when each interrupt occurs. To make the corresponding processing possible.

Task management 160 is responsible for sequentially executing the tasks generated by the system control 170 in accordance with their respective ranks and is in charge of the VRTX32 OS kernel.

The system control 170 performs overall management of the system, and creates and destroys tasks, manages the interrupt service routine 150, manages mailboxes and queues, manages variables, and the like.

The synthesis unit generation 180 converts the text received from the serial input / output unit 30 into parameters necessary for speech synthesis, and the generated parameters are transmitted to the speech synthesis unit 40 through DPRAM.

Meanwhile, the DSP unit 101 includes a DPRAM driver 190, a D / A conversion function 200, an interrupt service routine 210, and a PSOLA synthesis unit 220, and perform voice synthesis through actual signal processing.

The DPRAM driver 190 is used to exchange intermediate results with the central processing unit 2 and the intermediate parameters. The D / A converter 200 converts the digital voice synthesized by the timer interrupt every 0.625 msec and converts it into analog. Do this. The interrupt service routine 210 is responsible for processing interrupts generated by the DSP unit 101 and connecting the PSOLA synthesizer 220. The PSOLA synthesizer 220 generates the synthesized sound by connecting the waveform in the time domain using the parameters transmitted through the DPRAM, the rhythm rule, and the phonological knowledge, and the generated waveform is generated through the interrupt service routine 210. A converter 200 is passed.

4 is a structural diagram of a speech synthesis parameter according to the present invention, where N represents the number of synthesis units and a value of 1228. Each synthesis unit constituting the synthesis database is divided into six elements.

The synthesizing unit start position 310 indicates the starting position of the synthesizing unit in the database, and the segment symbol and the length 320 indicate the symbol representing the acoustic division in the synthesizing unit and the length of the symbol in the unit. The pitch start position 330 indicates the start position of the pitch data, and the pitch data 340 stores the pitch value of the synthesis unit. The sample start position 350 is a start position of the actual waveform, and the voice waveform is stored in the sample data 360.

5 is a flowchart of a speech synthesis processing according to the present invention, and is divided into a speech processing and a synthesis unit generation process performed by the central processing unit 20 and a speech synthesis performed by the speech synthesis unit 40.

The preprocessing unit 410 converts numbers and abbreviations into Korean, and the classifier 420 separates sentences through morphological analysis to generate syntax information, and the rhyme generator 430 generates 13 rhyme information. Generating information such as time, pitch, and the like, the letter-to-phonic converter 440 generates a phonetic phonetic sequence of sounds using 26 phonetic rules and exception dictionaries.

The synthesizing unit generator 450 converts each phonetic sequence into a synthesizing unit suitable for synthesis, and the synthesizing unit combiner 460 synthesizes the time domain required by the synthesizer by sequentially obtaining the parameters necessary for synthesis from a database. The PSOLA synthesizer 470, which generates the parameters and operates in the speech synthesizer 40, regularly adjusts the pitch, duration, and energy of the waveform in the time domain using the received parameters and phonetic knowledge. To produce

6 is an overall control flowchart of the text-to-speech service according to the present invention.

As shown in the figure, when a call occurs from the subscriber access unit 10, the central processing unit 20 attempts a corresponding call connection (510), and when the call connection is completed, the central processing unit 20 attempts to connect with the information provider. In step 515, when the connection with the information provider is completed, data received from the information provider is received (520), and data analysis is performed to analyze what stage it is at (525).

As a result of the analysis, if it is a step for requesting service termination, it is determined whether to release a termination message (530). At this time, if a DTMF signal corresponding to termination is received, the service provider disconnects and terminates the service (535). If not, it is analyzed whether the current state is a step of selecting a menu (540).

As a result of the analysis of the current state, if the menu step is a step, each menu and the guidance message is sent to the voice synthesis unit 40 to receive the menu data (545), and the Korean rule and the speech synthesis algorithm are applied to generate the synthesized sound (550). . The synthesized result is transmitted to the codec of the subscriber access unit 10 so that the user can hear it through the telephone (555).

On the other hand, if the result of the current state analysis is not the Matthew step, but the final search information desired by the user, data information detection for detecting the information section is performed (560), and a corresponding synthesized sound is generated (565) to the codec. The synthesized sound is transmitted (570).

After generating and transmitting a shout for the data transmitted from the information provider through the above process, the user receives the DTMF signal pressed by the user (575), and proceeds to the information provider through the data processor 30 to the next state. Send a command for (580). The process then returns to the process of receiving data from the information provider (520).

The present invention configured and carried out as described above can obtain the following special effects.

First, text information is converted into voice and provided to telephone subscribers, so anyone can get information quickly regardless of place.

Second, it is possible to expand the range of services that can be received from information providers not only users who have terminals but also telephone users.

Third, the user can listen to the desired information by voice, and can perform other tasks at the same time.

Fourth, since the information providing database consists only of letters, the information provider can reduce the burden of expense and time as compared to the case of building the database by recording method.

Fifth, information service can be provided to the visually impaired.

Sixth, it is possible to create a new information service using voice.

Claims (13)

Text-to-speech service applied to a communication network system comprising a telephone network connection unit 1 connected to a telephone network connected to a public telephone accommodating a general telephone subscriber and a packet network connection unit 3 connected between a public data network accommodating an information provider In the apparatus, detecting a ring signal and hook-on signal to send a hook-on signal to the telephone network connection unit (1) to request call disconnection, converts the synthesized analog voice into a PCM (Pulse Code Modulation) signal to the telephone network connection ( Subscriber access means 10 for transmitting to 1); Controlling the operation of the overall system, the central device configured to operate a high-speed device and a multitasking OS (Operating System) to perform a voice synthesis function for providing a synthesized voice to the subscriber access means 10 in real time Processing means 20; Data processing means (30) connected to the central processing means (20) and the packet network connection part (3) to perform data input / output with the outside; And the voice synthesis unit 40 converts a text into a voice, and synthesizes a digital voice under the control of the central processing unit 20 and transmits the synthesized voice to the subscriber access unit 10. Character-to-speech service device, characterized in that. The method of claim 1, wherein the subscriber access terminal 10 comprises: a codec for PCM signal conversion; A channel selector for setting a connection channel; And a DTMF converter for detecting a dual tone multifrequency (DTMF) signal and converting the signal to a predetermined bit. The CPU of claim 1, wherein the central processing unit 20 comprises a central processing unit (CPU), a floating point processing unit (FPU), a memory device (ROM, SRAM, DRAM, DPRAM), and a device controller. Character-to-speech service apparatus comprising a. 4. The memory device of claim 3, wherein the memory device comprises: a ROM having a system control function; a state RAM (SRAM) having a speech synthesis program requiring fast execution; A synchronous RAM (DRAM) having a synthetic database requiring a large memory; And a dual port RAM (DPRAM) used to exchange data with the voice synthesizing means (40). The data processing means (30) according to claim 1, wherein said data processing means (30) converts DTMF signals pressed by telephone subscribers into predetermined bit data and transmits the data to said packet network connection unit (3) and transmits data transmitted from an information provider to said central processing means ( A serial input / output unit transmitted to 20); And an Ethernet input / output unit used for debugging the system and downloading the synthesis database and the application program. 2. The voice synthesizing means (40) according to claim 1, characterized in that the voice synthesizing means (40) comprises a digital signal processor (DSP), a state RAM (SRAM), a ROM, an analog signal connection circuit (AIC), and a device controller. Text-to-speech service device. Subscriber access means 10 connected to an external telephone network connection 1, a central processing means 20 connected to the subscriber access means 10, between the central processing means 20 and an external packet network connection 3 In the text-to-speech service control method applied to a text-to-speech service apparatus having data processing means (30) connected to the terminal, when the call is generated from the subscriber access means (10), the central processing means (20) A first step of receiving data by connecting to an information provider through the packet network access unit (3); A second step of analyzing the received data to determine whether a current state is a step of selecting a menu unless the service is released; A third step of receiving menu data and generating synthesized sounds according to whether the menu selection step is performed, or generating and transmitting a corresponding synthesized sound by detecting data information; And a fourth step of receiving a DTMF signal pressed by the user and transmitting a command to the information provider to move to the next state through the data processing means (30). 8. The method of claim 7, wherein the second step comprises: a fifth step of analyzing whether the current data is analyzed to determine which step is present, and if the end of the service is requested, sending a termination message to confirm the release; And a sixth step of analyzing whether the current state is a step of selecting a menu when the DTMF signal corresponding to termination is received from the subscriber as a result of the checking, and disconnecting the information provider. Characterized by the voice-to-speech service control method. 8. The method of claim 7, wherein the third step further comprises: analyzing the menu selection step and, if the current state is the menu selection step, sending each menu and a guide message to the voice synthesizing means 40 to receive menu data. step; After performing the fifth step, applying a Korean rule and a speech synthesis algorithm to generate a synthesized sound, and transmitting it to the subscriber access means 10 so that the user can listen through the telephone; A seventh step of performing data information detection for detecting an information section when the search result is not the menu step but the final search information desired by the user as a result of analyzing the menu selection step; And an eighth step of generating a corresponding synthesized sound after performing the seventh step and transmitting the synthesized sound to the subscriber access means (10). The method of claim 7 or 9, wherein the process of generating the synthesized sound is divided into a language processing and synthesis unit generation process performed by the central processing means 20 and a speech synthesis process performed by the speech synthesis means 40. Character-to-speech service control method, characterized in that. The method of claim 10, wherein the process of generating a language and synthesizing unit comprises: a first process of converting numbers and abbreviations into Korean, classifying sentences through morphological analysis, and generating syntax information; A second step of generating information such as predetermined rhyme information, duration, pitch, and the like, and generating a phonetic symbol sequence using a predetermined phonological rule and an exception dictionary; And a third step of generating the synthesis parameters of the time domain by converting each phonetic symbol string into a synthesis unit suitable for synthesis and linking the parameters necessary for the synthesis from a database in order. How to control conversion services. The method of claim 11, wherein the speech synthesis process generates synthesized sound by regularly adjusting pitch, duration, and energy of the waveform in the time domain by using the received time domain synthesis parameters and phonetic knowledge. Character-to-speech service control method, characterized in that. The data structure of the synthesis parameter according to claim 11, further comprising: a synthesis unit start position field (310) indicating a start position of the synthesis unit in the database; A segment symbol and a length field 320 indicating a symbol indicating an acoustic division in the synthesis unit and a length of the symbol in the unit; A pitch start position field 330 indicating a start position of the pitch data; A pitch data field 340 in which pitch values of the synthesis unit are stored; A sample start position field 350 indicating a start position of an actual waveform; And a sample data field (360) in which a speech waveform is stored.