KR20200098320A - CTS conbine system with PU - Google Patents

Abstract

A CTS combine system with PU of the present invention comprises: a database unit in which a voice file is stored; a text input unit inputting a text; a control unit extracting a voice file corresponding to the text inputted from the text input unit from the database unit; a TTS engine converting the voice file transmitted from the control unit into audio wave data; and an audio processing unit connected to the TTS engine and converting the audio wave data converted in the TTS engine into an analog voice signal to transmit the analog voice signal to a speaker. The CTS combine system may prevent load.

Description

PU applied CTS conbine system {CTS conbine system with PU}

The present invention relates to a PU applied CTS Conbine system in which a character input by a character input through a personal terminal is matched with a voice.

The TTS (Text To Speech) system refers to a technology that converts a general text into a human voice. For example, a desired voice is generated by reading and combining the segmental sounds corresponding to each text stored in a database. This TTS system has been mainly applied to an unmanned automatic response system (ARS) or a technology that converts text information into voice and provides it for the visually impaired.

For example, in Korean Patent Registration No. 10-1180783, it is performed on a broadcasting terminal equipped with a function to convert text into speech (TTS: Text To Speech), but when user preference information is input, the user preference (A) updating the information database; Step (b) of constructing a broadcast program information database by receiving broadcast program information: When a customized TTS service is selected, the contents of the broadcast program information database and the taste information database are matched to create a customized broadcast program database suitable for the user's taste. Including (d) for converting the text data of the customized broadcasting program information database built in step (c) and (c) to be built by time or channel, and outputting it to voice, and provides a service of reservation viewing or reservation recording. A user-customized broadcast service method using the TTS technology further including the step (e) of instructing the terminal is proposed.

However, in the case of the above technology, there is no description of a problem in which an error or load is generated in the implementation of the TTS system after an error such as noise occurs in a voice file data converted into a database.

Korean Patent Registration No. 10-1180783

The present invention has been conceived to solve the above-described problems, and is intended to provide a PU applied CTS combine system capable of implementing a TTS system but increasing the degree of implementation by increasing the soundness of a voice file converted into data and preventing load.

To achieve the above object, the PU-applied CTS Convine system (hereinafter referred to as "system of the present invention") of the present invention transmits a character input unit for inputting a character, and a character signal input from the character input unit, and transmits a character signal to the character signal. A first communication unit that receives a matched voice file, a TTS engine that converts the voice file into audio wave data, and an audio processing unit that is connected to the TTS engine and converts the audio wave data converted by the TTS engine into an analog voice signal. And, a plurality of personal terminals including a speaker for expressing the voice signal converted by the audio processing unit; And a central server including a second communication unit for receiving the text signal and transmitting a voice file matching the text signal, and a database unit storing the voice file.

For example, in the central server, a voice input unit for inputting a voice file to the database unit is configured, and the voice input unit includes a near-field microphone at a distance from the input person, a remote microphone at a distance from the input person, and the near microphone and the remote The frequency component of the input signal input from the microphone is calculated through Fourier transform, the frequency signal power is calculated through the frequency component of each microphone, and the frequency signal power of the near-field microphone and the frequency signal power of the far-field microphone are compared and analyzed to obtain both signals. And a noise canceling unit that calculates a power ratio and determines whether or not there is a voice file to be stored in the database based on the calculated ratio of the power of both signals.

As an example, the voice input unit includes a housing to which the near microphone and the far microphone are exposed, and the noise canceling unit is embedded in the housing, and the far microphone has a plurality of flow holes in the direction of the near microphone, and a closed surface is formed in the opposite direction. It is characterized in that it is configured in a shape covered by a reflective wave blocking cover to be formed.

As an example, the voice input unit includes a housing to which the near-field microphone and the far-field microphone are exposed, and in the housing, a concave-shaped guide groove is formed in the far-field microphone, and the far-end microphone is the end of the guide groove. It is formed on the side wall and characterized in that the reflection wave blocking cover is formed at the end of the guide groove on which the long-distance microphone is formed.

The system of the present invention is applied to a variety of uses, such as an automatic response system for announcements, public transport guidance navigation, and educational applications, and has advantages in that effects such as convenience, consistency of voice, and economy can be expressed.

In addition, there is an advantage of increasing the operational efficiency of the TTS system by allowing only pure audio files from which noise has been removed to be stored as data when storing audio files.

1 is a schematic diagram showing the system of the present invention.
Figure 2 is a block diagram showing a detailed configuration of the personal terminal in Figure 1;
Figure 3 is a block diagram showing a detailed configuration of the central server in Figure 1;
4 is a block diagram showing a detailed configuration of a voice input unit, which is one configuration of the present invention.
5 and 6 are schematic diagrams showing each embodiment of the voice input unit.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings.

The system 1 of the present invention includes a character input unit 21 for inputting characters as shown in Figs. 1 to 3, and a voice file matching the character signal by transmitting a character signal input from the character input unit 21. The first communication unit 25 receiving a signal, a TTS engine 23 that converts the voice file into audio wave data, and the audio wave data that is connected to the TTS engine 23 and converted by the TTS engine 23 A plurality of personal terminals (2) including an audio processing unit (24) for converting into an analog audio signal and a speaker (26) for expressing the audio signal converted by the audio processing unit (24); And a central server (3) including a second communication unit (34) for receiving the text signal and transmitting a voice file matching the text signal, and a database unit (32) storing the voice file. .

The personal terminal 2 includes a character input unit 21, a terminal control unit 22, a TTS engine 23, an audio processing unit 24, a first communication unit 25, and a speaker 26 as shown in FIG. It characterized in that it is included.

In addition, the central server 3 is characterized in that it includes a voice input unit 31, a database unit 32, a central control unit 33, and a second communication unit 34 as shown in FIG.

Hereinafter, an operation mechanism of the above components will be described.

The character input unit 21 is connected to the terminal control unit 22 and includes a plurality of numeric keys and function keys for performing various functions.

The terminal control unit 22 transmits the character signal inputted from the character input unit 21 to the first communication unit 25 so that the first communication unit 25 transmits the character signal to the central server 3 Is to do.

The text signal is received by the second communication unit 34 of the central server 3, and the central control unit 33 extracts a voice file matching the received text signal from the database unit 32, and the second communication unit ( It is to transmit to the first communication unit 25 of the personal terminal 2 through 34).

When a voice file is received from the first communication unit 25 in this way, the terminal control unit 22 transmits the received voice file to the TTS engine 23, and the TTS engine 23 transmits the received voice file to the terminal control unit 22. It is connected to one side of the terminal control unit 22 to convert a text sentence inputted in a predetermined language, that is, a voice file into audio wave data.

The audio processing unit 24 is connected to one side of the TTS engine 23 and converts the audio wave data converted by the TTS engine 23 into an analog audio signal. The audio processing unit 24 is a general software module and includes an audio driver and an audio card as a hardware block.

In the central server 3, the database unit 32 stores various voice files and corresponds to a configuration in which voice files input by the voice input unit 31 to be described below are stored.

In addition, as shown in FIG. 3, in the present invention, a voice input unit 31 for inputting a voice file to the central server 3 is configured. In particular, in the present invention, noise is generated in the voice file converted into data in the database unit 32. An embodiment for solving a problem in which an error occurs or a load is generated in the implementation of the TTS system is proposed.

As shown in FIG. 4, the voice input unit 31 of the present embodiment includes a near-field microphone 311 at a distance from the input user, a far-field microphone 312 at a distance from the input user, and the near-field microphone 311 and the far-field microphone ( Calculate the frequency component through Fourier transform of the input signal input from 312), calculate the frequency signal power through the frequency component of each microphone, and calculate the frequency signal power of the near-field microphone 311 and the frequency signal of the far-field microphone 312 And a noise reduction unit 313 that compares and analyzes power to calculate a ratio value of both signal powers and determines whether there is a voice file to be stored in the database unit 32 based on the calculated ratio value of both signal powers. To do.

The near-field microphone 311 is positioned close to the inputter's mouth, and the far-field microphone 312 is positioned relatively far from the input's mouth compared to the near-field microphone 311.

The noise reduction unit 313 first converts an analog signal input from each of the microphones 311 and 312 into a digital signal through an analog to digital converter.

The noise removal unit 313 compares and analyzes the frequency signal power of the near-field microphone 311 and the frequency signal power of the far-field microphone 312 to calculate a ratio of both signal powers, and based on the calculated ratio of the two signal powers. Thus, the presence or absence of a voice file to be stored in the database unit 32, that is, signals other than the voice file are estimated as noise and removed.

In more detail, the noise canceling unit 313 identifies the frequency components of the voice signal of the near-field microphone 311 and the voice signal of the far-field microphone 312 input in the form of a digital signal. Fourier transform is performed on. At this time, if the input voice signal of the near-field microphone 311 is d(n) and the input voice signal of the far-field microphone 312 is x(n), d(k) by Fourier transform of d(n) And x(k) by the Fourier transform of x(n) are derived by Equation 1 below.

Here,'N' is the number of samples of the block during the previous predetermined period including the current sample during Fourier transform, and is 0≦n≦N-1 and 0≦k≦N-1.

Thereafter, the noise removing unit 313 calculates the frequency signal power of both microphones 311 and 312, and calculates the frequency signal power of the k-th frequency components d(k) and x(k) in Equation 1 above. Considering D(k) and X(k), respectively, the D(k) and X(k) are derived by Equation 2 below.

As described above, when the two microphones 311 and 312 are located close together, in the case of a noise signal that is relatively far apart from the input's mouth, the noise signal is input to each microphone 311 and 312 with almost the same signal power. do.

On the contrary, the voice of the inputter is input to the near-field microphone 311 close to the input person's mouth with a higher level of frequency signal power than the far-field microphone 312.

In the present embodiment, by estimating the noise section using this phenomenon, only the voice file of authenticity is converted into data in the database unit 32 so as to improve the implementation efficiency of the TTS system.

First, a method for estimating a noise section according to an embodiment of the present invention will be described when D(k) <X(k) in the D(k) and X(k) values derived in Equation 2 above, that is, the distance When the frequency signal power of the microphone 312 is greater than the frequency signal power of the near-field microphone 311, the signal power D(k) of the two microphones 311, 312 in the lth frame, the ratio value A(l) of X(k) ) Is derived by Equation 3 below.

Here, N is the number of samples in one block, and l is a frame index.

In general, since noise is input to a microphone at a relatively farther distance than the input's mouth, a frequency signal power of almost similar level can be input to the two microphones 311 and 312, and the A(l) value at this time is almost "1. "It gets closer to.

Accordingly, the noise removal unit 313 according to the present embodiment estimates the noise interval by comparing and analyzing the ratio values A(l) of D(k) and X(k) derived from Equation 3 with a threshold value. , If A(l)>Thr_A is satisfied in the l-th frame, it can be estimated as a noise interval. Here, the threshold value Thr_A is a value between 0 and 1.

The threshold value Thr_A is an optimal set value obtained from repeated experiment results, and the present invention is not limited thereto.

As a method for estimating a noise section according to another embodiment, in the case of D(k)>X(k) in the values of D(k) and X(k) derived in Equation 2, as opposed to the above-described embodiment, That is, when the frequency signal power of the near-field microphone 311 is greater than the frequency signal power of the far-field microphone 312, the ratio B of the power signals D(k) and X(k) of the two microphones 81 and 82 in the lth frame (l) is derived through Equation 4 below.

Here, N is the number of samples in one block, and l is a frame index.

In this embodiment, since the input's voice is input at a relatively higher level to the near-field microphone 311 than in the far-field microphone 312, B(l) at this time is less than “1” and A ( l).

Accordingly, B(l) in the l-th frame can be estimated as a noise interval when B(l) &lt; Thr_B is satisfied. Here, the threshold value Thr_B is a value between 0 and 1. Here, the threshold value Thr_B is also an optimal set value obtained from an iterative experiment result, and the present invention is not limited thereto.

The B(l) value represents a large value in the section without voice activity (Noise), and a relatively small value in the section with voice activity. Noise is removed by extracting the spectrum of the noise signal only in the section without voice activity. To do it.

Through the above-described two embodiments, the noise removing unit 313 separates the voice file section and the noise section, and can extract and remove only the noise signal.

Meanwhile, according to a preferred embodiment of the present invention, in order to protect the voice file section when estimating the noise section by the noise removing unit 313, A(l) and B derived from Equations 3 and 4 are (l) Perform smoothing for each.

For example, for A(l)

i) When A(l)≥A(l-1), A(l)=α1*A(l)+(1-α1)*A(l-1)

ii) When A(l)<A(l-1), A(l)=α2*A(l)+(1-α2)*A(l-1)

Here, 0<α1<α2<1.

When smoothing as described above, A(l) increases slowly and decreases quickly.

Conversely, in the case of B(l)

i) When B(l)>B(l-1), B(l)=β1*B(l)+(1-β1)*B(l-1)

ii) When B(l)≤B(l-1), B(l)=β2*B(l)+(1-β2)*B(l-1)

Here, 1>β1>β2>0.

When smoothing as described above, B(l) increases relatively faster than A(l) and decreases slowly.

Here, the smoothing method may be variously implemented through known techniques, but is not limited to the examples presented in the present invention.

Thereafter, when the ratio value of the power of both signals is not estimated as a noise section through a comparison analysis with a threshold value, the noise removal unit 313 determines that there is voice activity and outputs a signal indicating that there is voice activity. That is, it is determined as a voice file section.

The signal and the voice file signal estimated as such a noise section are input from each microphone 311 and 312 based on the signal input from the noise reduction unit 313 and selectively edit the signal from the digitally converted signal, The noise is removed.

As described above, in the present invention, noise that may be input to the database unit 32 is eliminated through the above operation mechanism. These noises are respectively used by the microphones 311 and 312 together with the input's voice signal, the noise of the computer fan in the room, and the TV. As it is introduced as ambient noise such as sound, it seems that this noise can be easily removed by the operation mechanism of the noise reduction unit 313, but in the case of a reflected wave formed by reflecting a voice signal to furniture, etc., it has a frequency similar to that of the voice signal. There is a problem that it is not easy to remove this in the noise removing unit 313. That is, in the case of the reflected wave, the noise canceling unit 313 may be a factor in which an incorrect voice file may be converted into data.

Accordingly, in the present invention, two embodiments are proposed, and the first embodiment is shown in FIG. 3.

The voice input unit 31 of the present embodiment includes a housing 314 to which the near-field microphone 311 and the far-field microphone 312 are exposed, and the noise canceling unit 313 is embedded in the housing 314, and The microphone 312 has a shape covered by a reflection wave blocking cover 315 forming a closed surface 315-2 in the opposite direction and a plurality of flow holes 315-1 formed in the direction of the near-field microphone 311 It features.

The far-field microphone 312 is configured in a state embedded in the reflected wave blocking cover 315, and the input is close to the near-field microphone 311 to generate a voice signal, and this voice signal S1 is the far-field microphone 312 ) Is introduced through the flow hole 315-1, and in the case of the reflected wave for the voice signal, the reflected wave is transmitted to the far-field microphone 312 by the closed surface 315-2 of the reflected wave blocking cover 315. It is to prevent inflow.

That is, the near-field microphone 311 inputs the input's voice with a higher level of frequency signal power than the far-field microphone 312, whereas the far-field microphone 312 inputs the input's voice with a small level of frequency signal power. Since the risk of signal distortion is great, it is to block the inflow of the reflected wave from the far-field microphone 312.

Preferably, the reflected wave blocking cover 315 is made of various known sound absorbing materials so that the reflected wave is absorbed by the reflected wave blocking cover 315 is reasonable.

A second embodiment is shown in FIG. 6.

The voice input unit 31 according to the present embodiment includes a housing 314 to which the near-field microphone 311 and the far-field microphone 312 are exposed, and in the housing 314, the near-field microphone 311 to the far-field microphone ( In 312, a concave-shaped induction groove 316 is formed, and the distance microphone 312 is formed on an end sidewall of the induction groove 316, and a reflection wave is formed at the end of the induction groove 316 in which the distance microphone 312 is formed. It is characterized in that the blocking cover 317 is formed.

The far-field microphone 312 is configured in a state embedded in the induction groove 316 and the reflected wave blocking cover 317, and the inputter generates an audio signal in proximity to the near-field microphone 311, such a voice signal (S1) In the case of direct flow into the far-field microphone 312, the reflected wave (S2) for the voice signal is transmitted through the flow path formed by the guidance groove 316 and the reflected wave blocking cover 317 through the guidance groove 316. In this case, the inflow of the reflected wave to the far-field microphone 312 is blocked by the reflected wave blocking cover 317.

In this case, the input's voice is input to the near-field microphone 311 with a higher level of frequency signal power than the far-field microphone 312, whereas the far-field microphone 312 receives the input's voice with a small level of frequency signal power. Since the risk of signal distortion is large due to the influence, it is to block the inflow of the reflected wave from the far-field microphone 312.

Also in the case of the present embodiment, it is appropriate that the reflected wave blocking cover 317 uses various known sound absorbing materials so that the reflected wave is absorbed by the reflected wave blocking cover 317.

It will be appreciated by those skilled in the art through the above description that various changes and modifications can be made without departing from the technical idea of the present invention. Therefore, the technical scope of the present invention should not be limited to the contents described in the detailed description of the specification, but should be determined by the claims.

1: the present invention 2: personal terminal
3: Central server

Claims (4)

A character input unit for inputting a character, a first communication unit for transmitting a character signal input from the character input unit and receiving a voice file matching the text signal, a TTS engine for converting the voice file into audio wave data, and the A plurality of personal terminals including an audio processing unit connected to the TTS engine and converting audio wave data converted by the TTS engine into an analog audio signal, and a speaker generating the audio signal converted by the audio processing unit; And
And a central server including a second communication unit for receiving the text signal and transmitting a voice file matched with the text signal, and a database unit storing the voice file.
The method of claim 1,
In the central server,
A voice input unit for inputting a voice file to the database unit is configured, and the voice input unit Fourier inputs an input signal input from the input person to the near-field microphone, the far-field microphone and the near-field microphone and the far-field microphone. The frequency component is calculated through conversion, the frequency signal power is calculated through the frequency component of each microphone, and the ratio of the power of both signals is calculated by comparing and analyzing the frequency signal power of the near-field microphone and the frequency signal power of the far-field microphone. And a noise canceling unit for determining the presence or absence of a voice file to be stored in the database unit based on a ratio value of the power of both signals.
The method of claim 2,
The voice input unit includes a housing to which the near-field microphone and the far-field microphone are exposed, and the noise-removing portion is embedded in the housing, and the far-field microphone has a plurality of flow holes in the near-field microphone direction, and a reflected wave forming a closed surface in the opposite direction. PU applied CTS Conbine system, characterized in that it is configured in a shape covered by a blocking cover.
The method of claim 2,
The voice input unit includes a housing to which the near-field microphone and the far-field microphone are exposed, and in the housing, a concave-shaped induction groove is formed in the far-field microphone in the near-field microphone, and the far-field microphone is formed on a side wall of the end of the guide groove. And a reflective wave blocking cover is formed at the end of the guide groove on which the far-field microphone is formed.

Images