KR101961341B1 - Signal processing apparatus and method for barge-in speech recognition - Google Patents

Abstract

[0001] The present invention relates to an apparatus and method for signal processing for pants-in speech recognition, in which an attenuated analog gain value is stored so that clipping does not occur with respect to an output volume value of a speaker where clipping of echo input through a microphone occurs, It is necessary to check the corresponding analog gain value according to the output volume value of the speaker outputting the voice, amplify the analog signal and convert it into a digital signal to remove the echo, By amplifying the signal, it is possible to effectively remove the echo mixed with the uttered voice of the user, and the accuracy of speech recognition is improved.

Description

[0001] The present invention relates to a signal processing apparatus and method for speech recognition,

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a signal processing technique for speech recognition, and more particularly to a signal processing technique for processing a signal inputted through a microphone for voice recognition in a barge- To a signal processing apparatus and method for voice recognition of pants-in.

Speech recognition technology refers to a technique of converting an acoustical signal according to a user's uttered voice into words or sentences. Generally, in a system for providing a voice recognition function, a user hears a prompt sound outputted from a speaker to proceed with speech recognition, utterances the voice corresponding thereto, and inputs the voice to the microphone. At this time, So that it is permitted to ignite only after it is finished, which is inconvenient for the user who uses the speech recognition function.

In order to improve the inconvenience, a speech recognition technology using a barge-in function is being developed. A barge-in is a function that allows a user to utter a voice before completion of output of a guidance voice in a voice recognition system in which a guidance voice for voice recognition is output to a speaker.

In the speech recognition system, when the user uses the pants-in function, the user does not have to wait until the guidance voice is finished and can understand the meaning according to the partial output of the guidance voice and immediately utter the voice to be input. Time can be shortened.

However, when the guidance voice is generated by the pants-in method during the output of the guidance voice, the guidance voice output from the speaker is micro-input together with the user's speaking voice in the form of echo. Can be improved.

However, when clipping occurs in the echo input through the microphone due to the level of the guidance voice output from the speaker, it is difficult to effectively remove the echo, so a method for improving the echo is required.

Published Japanese Patent Application No. 10-2003-0073886 (published September 19, 2003)

SUMMARY OF THE INVENTION It is an object of the present invention to solve the above problems and to provide a method and apparatus for detecting a voice of a guidance voice input through a microphone in the form of an echo during speech recognition using a barge- The present invention is to provide a signal processing apparatus and method for voice recognition of pants-in which clipping is prevented, echo can be effectively removed, and voice recognition efficiency can be enhanced.

According to another aspect of the present invention, there is provided a signal processing apparatus for voice recognition of a pants-in, comprising: a storage unit for storing analog gain value information of a microphone corresponding to an output volume value of a speaker for outputting a guidance voice; A first amplifier for amplifying an analog signal input through the microphone according to an analog gain value stored in the first amplifier, an analog-to-digital converter for converting the analog signal amplified by the first amplifier into a digital signal, A second amplifier for amplifying the digital signal from which the echo is removed, and an analog gain value information of the microphone corresponding to the output volume value of the speaker, Wherein the clipping of the echo inputted through the microphone occurs in the speaker Stores the attenuated analog gain value so that clipping does not occur with respect to the output volume value, and confirms the corresponding analog gain value in the storage unit according to the output volume value of the speaker outputting the guidance voice, And a controller for controlling the first amplifier to amplify the analog signal inputted through the microphone according to the analog gain value and for controlling the second amplifier to recover the digital signal as much as the analog signal is attenuated.

In the signal processing device for pants-in speech recognition of the present invention, when the guidance sound is output from the speaker, the controller checks an output volume value at which clipping occurs in the echo input through the microphone, Calculating an attenuated analog gain value such that clipping does not occur by reducing the analog gain value corresponding to the output volume value, and storing the clipped-confirmed output volume value and the corresponding attenuated analog gain value in the storage section .

In the signal processing apparatus for speech recognition of the present invention, the control unit calculates an analog gain value corresponding to the output volume value of the speaker using the following equation and stores it in the storage unit .

인 경우

If

인 경우

If

Where AG is the analog gain value, v is the output volume value, T is the maximum volume value at which no clipping occurs, A is the non-attenuated analog gain value, and S is the output level of the speaker.

The controller calculates a digital gain value for restoring a digital signal as much as the analog signal is attenuated in the first amplifier by using the following equation, And controls the second amplifier to amplify the digital signal according to the calculated digital gain value.

인 경우

If

인 경우

If

In this case, DG represents a digital gain value, AG represents an analog gain value, v represents an output volume value, T represents a maximum volume value at which clipping does not occur, and A represents an attenuated analog gain value.

The apparatus further comprises a speech recognizer for recognizing a speech sound inputted together with the echo through the microphone using the digital signal amplified by the second amplifier, do.

According to another aspect of the present invention, there is provided a signal processing method for voice recognition of a pants-in-speech, the method comprising: storing information on analog gain of a microphone corresponding to an output volume value of a speaker for outputting a guidance voice, Storing an attenuated analog gain value such that clipping does not occur with respect to an output volume value of the speaker where clipping of an echo input through the microphone occurs; Determining a corresponding analog gain value, amplifying an analog signal inputted through the microphone in accordance with an identified analog gain value, the signal processing apparatus converting the amplified analog signal into a digital signal The signal processing apparatus comprising: Removing an echo, and the signal processing unit corresponding to the to recover the digital signal by the analog signal attenuation includes the step of amplifying the digital signal of the echo is eliminated.

In the signal processing method for voice recognition of a pants-in voice of the present invention, when the signal processing apparatus outputs a guidance voice from the speaker, an output volume value at which clipping occurs in the echo input through the microphone Calculating a decayed analog gain value such that clipping does not occur by reducing the analog gain value corresponding to the output volume value at which the signal processing apparatus is confirmed to be clipped; and calculating, by the signal processing apparatus, Value and a corresponding attenuated analog gain value.

In the signal processing method for voice recognition of the present invention, the signal processing apparatus may calculate and store an analog gain value corresponding to the output volume value of the speaker using the following equation .

인 경우

If

인 경우

If

Where AG is the analog gain value, v is the output volume value, T is the maximum volume value at which no clipping occurs, A is the non-attenuated analog gain value, and S is the output level of the speaker.

In the signal processing method for voice recognition of a pants-in of the present invention, the step of amplifying may include: calculating a digital gain value for recovering a digital signal as much as the analog signal is attenuated using the following equation And amplifies the echo-canceled digital signal by calculation.

인 경우

If

인 경우

If

In this case, DG represents a digital gain value, AG represents an analog gain value, v represents an output volume value, T represents a maximum volume value at which clipping does not occur, and A represents an attenuated analog gain value.

In the signal processing method for voice recognition of pants in accordance with the present invention, the signal processing apparatus further comprises a step of recognizing a speech sound inputted together with the echo through the microphone using the amplified digital signal .

According to another aspect of the present invention, there is provided a computer-readable recording medium having recorded thereon a program for performing a signal processing method for voice recognition.

According to the signal processing apparatus and method for voice recognition of the pants-in of the present invention, when the voice of the user is uttered during the output of the guidance voice for voice recognition, and the voice of the user and the guidance voice of the echo- It is possible to effectively remove only the corresponding signal.

Especially, even when the level of the guidance voice output from the speaker is high and clipping may occur in the echo inputted through the microphone, clipping can be prevented by appropriately adjusting the gain of the signal inputted through the microphone, Thus, the speech recognition efficiency can be improved by preventing the speech speech distortion.

As described above, according to the present invention, the performance of the pants-in speech recognition can be improved by using the analog gain of the microphone differently according to the speaker output level. In the range of the low speaker output level, And at a high speaker output level, a low analog gain value is used to prevent the echo being input to the microphone from being clipped, thereby effectively eliminating echoes and preventing spurious distortion.

In this way, it is possible to remove the micro input echo using a small analog gain value regardless of the speaker output volume value, and then amplify according to the digital gain value to restore the size of the user's utterance signal. It is possible to prevent the loss of the digital representation value of the user's utterance signal that occurs in the system.

1 is a block diagram of a signal processing apparatus according to an embodiment of the present invention.
2 is a diagram illustrating output signals of a speaker according to an exemplary embodiment of the present invention.
3 is a diagram illustrating echoes input through a microphone according to an embodiment of the present invention.
4 is a diagram illustrating clipping occurring in an echo input through a microphone according to an embodiment of the present invention.
5 is a diagram illustrating analog gain value information of a microphone corresponding to an output volume value of a speaker according to an embodiment of the present invention.
6 is a diagram illustrating a process of calculating a gain value for signal processing according to an embodiment of the present invention.
7 is a diagram illustrating a process of processing a signal according to the embodiment of FIG.

In the following description, only parts necessary for understanding the embodiments of the present invention will be described, and the description of other parts will be omitted so as not to obscure the gist of the present invention.

The terms and words used in the present specification and claims should not be construed as limited to ordinary or dictionary meanings and the inventor is not limited to the meaning of the term in order to describe his invention in the best way. It should be interpreted as meaning and concept consistent with the technical idea of the present invention. Therefore, the embodiments described in the present specification and the configurations shown in the drawings are merely preferred embodiments of the present invention, and are not intended to represent all of the technical ideas of the present invention, so that various equivalents And variations are possible.

The present invention relates to the field of signal processing for speech recognition. Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a diagram illustrating a configuration of a signal processing apparatus according to an embodiment of the present invention. FIG. 2 is a diagram illustrating output signals of a speaker according to an embodiment of the present invention. FIG. FIG. 4 is a diagram illustrating a clipping occurring in an echo input through a microphone according to an embodiment of the present invention. FIG. 5 is a diagram illustrating an echo input through a microphone according to an embodiment of the present invention. And the analog gain value information of the microphone corresponding to the output volume value of the speaker.

1 to 5, the signal processing apparatus 100 according to the present embodiment is configured such that a voice uttered by the user 2 according to a guidance voice prompt output from the speaker 1 is transmitted to a microphone 3 And then processes the input signal for voice recognition.

First, when a guidance voice is output from the speaker 1 for voice recognition, the user 2 comprehends the contents of the guidance voice and then utteres the corresponding voice to input the voice to the microphone 3. At this time, the user 2 can input a speech sound into the microphone 3 by using a barge-in function that allows speech during speech while outputting a guidance voice through the speaker 1.

At this time, the guidance voice output from the speaker 1 is input to the microphone 3 in the form of echo and is input to the microphone 3 in a form mixed with the voice uttered by the user 2. The signal processing device 100 Echo canceller technology is used to remove the echo due to the guidance voice among the input signals through the microphone 3 and to extract only the speech voice of the user 2 for speech recognition, have.

The signal processing apparatus 100 includes a storage unit 10, a first amplifier 20, an analog-to-digital converter 30, an echo canceller 40, a second amplifier 50, a voice recognizer 60, ).

The storage unit 10 stores analog gain value information of the microphone 3 corresponding to the output volume value of the speaker 1 that outputs the guidance voice, and includes various memories for storing information. The information stored in the storage unit 10 is referred to when the control unit 70 controls the amplification operation of the first amplifier 20. The analog gain value stored in the storage unit 10 may have different gain values depending on the output volume value of the speaker 1. [

The first amplifier 20 amplifies the analog signal input through the microphone 3 in response to the analog gain value stored in the storage unit 10 under the control of the controller 70. At this time, the signal input through the microphone 3 may be a signal in which a guidance voice output through the speaker 1 and a speech voice of the user 2 are mixed according to a barge-in function.

An analog-to-digital converter (ADC) 30 converts the analog signal amplified by the first amplifier 20 into a digital signal.

The echo canceller 40 serves to remove an echo signal corresponding to the announcement voice from the digital signal converted by the analog-to-digital converter 30.

The echo canceller 40 removes the echo included in the signal input through the microphone 3 by using a signal output from the speaker 1 as a reference. The echo canceller 40 uses an adaptive filter to remove the echo from the speaker 1 (Echo estimate), which is the same signal as the echo using the estimated value and the reference signal, after estimating a time delay and an amplitude change from when the guidance voice is outputted to the microphone 3, ) Signal. The echo canceller 40 subtracts the echo estimation signal from the input signal through the microphone 3 and removes the signal corresponding to the echo.

The second amplifier 50 amplifies the echo-canceled digital signal under the control of the controller 70.

The voice recognizer 60 recognizes the voice of the user 2 inputted together with the echo via the microphone 3 by using the digital signal amplified by the second amplifier 50. [ The signal processing apparatus 100 may process an input signal through the microphone 3 and transmit the processed signal to an external speech recognizer 60. Alternatively, It is possible to perform speech recognition directly using the speech recognizer 60. [

The control unit 70 includes a storage unit 10, a first amplifier 20, an analog-to-digital converter 30, an echo canceller 40, a second amplifier 50, 100, and includes an operation unit, a memory, a program storage, and the like for this purpose.

First, the control unit 70 stores the analog gain value information of the microphone 3 corresponding to the output volume value of the speaker 1 in the storage unit 10.

The output volume value of the speaker 1 is proportional to the output level of the guidance voice output from the speaker 1. When the output volume value of the speaker 1 is too large, the level of the echo is also very large, And the echo signal transmitted through the digital-analog amplifier 30 for voice recognition is saturated beyond a limit that can be expressed, and clipping occurs at a maximum value or a minimum value.

FIG. 2 shows a guidance voice output from the speaker 1 when the resolution of the digital-analog amplifier 30 is 16 bits. FIG. 3 shows a configuration in which a guide voice of an echo type is input to the microphone 3 . In this case, since the output level of the speaker 1 is not large, clipping does not occur.

On the other hand, FIG. 4 shows a state in which the output level of the speaker 1 is clipped. When the resolution of the digital-analog amplifier 30 is 16 bits, the maximum value is 32767 and the minimum value is -32768. When the output volume value of the speaker 1 is too large, saturation exceeding the range occurs, Or clipping occurs within the range of the minimum value, resulting in nonlinear distortion. As a result, much information of the signal disappears, and it becomes difficult to remove the echo from the echo canceller 40, and when the echo is removed by the post processing technique, distortion of the voice of the user 2 overlapped with the echo .

For this purpose, the control unit 70 stores the analog gain value information of the microphone 3 corresponding to the output volume value of the speaker 1 in the storage unit 10, And stores the attenuated analog gain value in the storage unit 10 so that clipping does not occur with respect to the output volume value of the speaker 1 where clipping of the echo inputted through the microphone 3 occurs.

At this time, if the guidance sound is outputted from the speaker 1 before the user 2 is ignited, the control unit 70 checks whether clipping occurs in the echo inputted to the microphone 3, and if the clipping occurs, the control unit 70 confirms the output volume value . And the attenuated analog gain value is calculated so that clipping does not occur by reducing the analog gain value corresponding to the detected output volume value. At this time, the attenuated analog gain value has a relatively smaller value than the analog gain value corresponding to the clipping-free output volume value, and the controller 70 outputs the clipped-confirmed output volume value and the corresponding attenuated analog gain value Is stored in the storage unit (10) and referred to the amplification operation control of the first amplifier (20).

In this case, the controller 70 calculates the analog gain value corresponding to the output volume value of the speaker 1 using the following equation and stores it in the storage unit 10.

인 경우

If

In this case, AG represents the analog gain value, v represents the output volume value, T represents the maximum volume value at which clipping does not occur, A represents the non-attenuated analog gain value, and S represents the output level of the speaker (1).

The table shown in FIG. 5 shows the analog gain value information stored in the storage unit 10 by the controller 70 according to the calculated result. The output volume value of the speaker 1 is located in the left column, The output power of the speaker 1 is located in the center column, and the analog gain value information of the microphone 3 corresponding to each output volume value is located in the right column.

5, when the output volume value of the speaker 1 is 1 to 7, the echo input through the microphone 3 is not saturated and is not attenuated, and represents a general state analog gain value A . On the other hand, when the output volume value of the speaker 1 is 8 to 10, since the echo inputted through the microphone 3 is saturated and clipping occurs, the attenuated analog gain value ( B, B, D).

According to Equation (1), for example, the analog gain value corresponding to the output volume value 8 of the speaker 1 in FIG. And the maximum volume value at which clipping does not occur is the maximum value of the volume value having the analog gain value A, so T has a value of 7. Since the output level of the speaker 1 at the output volume value 7 is S (7) and the output level of the speaker 1 at the output volume value 8 is S (8), the output level of the speaker 1 at the output volume value 8 The corresponding analog gain value B is determined as follows.

B? A - [S (8) - S (7)] dB

At this time, the output level of the speaker 1 when the output volume value is 8 is relatively large compared to the output level of the speaker 1 when the output volume value is 7 (S (8) - S (7) (8) -S (7)] as the output of the speaker 1 becomes larger at a general analog gain value A, the level of the echo at the output volume value 8 is clipped It can be maintained at a level that does not.

Similarly, the analog gain value C corresponding to the output volume value 9 of the speaker 1 is determined as follows.

C? A - [S (9) - S (7)] dB

The analog gain value D corresponding to the output volume value 10 of the speaker 1 is determined as follows.

D? A - [S (10) - S (7)] dB

After the control unit 70 stores the analog gain value information of the microphone 3 corresponding to the output volume value of the speaker 1 in the storage unit 10 and then outputs the guidance voice from the speaker 1, Confirms the output volume value of the speaker 1 that outputs the guidance voice, and then confirms the corresponding analog gain value by referring to the information stored in the storage unit 10. [ The control unit 70 controls the first amplifier 20 according to the identified analog gain value to amplify the analog signal input through the microphone 3. As a result, no clipping occurs in the signal amplified by the first amplifier 20.

The amplified analog signal is converted into a digital signal by the analog-to-digital converter 30 so that clipping does not occur in the first amplifier 20, and the echo corresponding to the guidance voice is removed from the echo canceler 40.

Thereafter, the controller 70 controls the second amplifier 50 to amplify the echo-canceled digital signal. At this time, the amplification operation of the second amplifier 50 is performed to recover the digital signal as much as the analog signal is attenuated, .

In this case, the control unit 70 calculates a digital gain value for restoring the digital signal as much as the analog signal is attenuated in the first amplifier 20 by using the following equation, and outputs the digital gain value to the second amplifier 50 to amplify the digital signal.

인 경우

If

In this case, DG represents the digital gain value, AG represents the analog gain value, v represents the output volume value, T represents the maximum volume value at which clipping does not occur, and A represents the non-attenuated analog gain value.

That is, if the current output volume value of the speaker 1 for outputting the guidance voice is equal to or less than T, which is the maximum volume value at which clipping does not occur, the controller 70 sets the analog gain value referenced when controlling the first amplifier 20 to Since there is no attenuation, the digital gain value is calculated as 0 so that the second amplifier 50 does not amplify the digital signal.

On the other hand, if the current output volume value of the speaker 1 outputting the announcement voice exceeds the maximum volume value T that does not cause clipping, the control unit 70 determines that the analog gain referred to when controlling the first amplifier 20 The second amplifier 50 calculates a digital gain value so as to amplify the digital signal by an amount corresponding to the attenuation of the analog signal and controls the second amplifier 50 according to the calculated digital gain value Amplify.

At this time, the signal amplified by the second amplifier 50 is transmitted to the voice recognizer 60 as an echo-canceled signal, and is used to recognize the voice.

At this time, the voice recognizer 60 may be included in the signal processing apparatus 100, and may be separately provided in some embodiments. The controller 70 transmits the digital signal amplified by the second amplifier 50 to the external speech recognizer 60 so that speech recognition is performed Can be supported.

The process of processing a signal for speech recognition in accordance with the present invention will be described in detail with reference to FIG. 6 and FIG.

6 is a diagram illustrating a process of calculating a gain value for signal processing according to an embodiment of the present invention.

Referring to FIG. 6, a signal processing device for pants-in speech recognition recognizes an output volume value at which clipping occurs in an echo input through a microphone (S2) when a guidance voice is output from a speaker (S1).

Then, the signal processing apparatus calculates an attenuated analog gain value so as to prevent clipping by reducing the analog gain value corresponding to each output volume value of the clipped speaker output volume (S3) Value and a corresponding digital gain value for restoring the attenuated analog gain value (S4).

In step S3, the signal processing apparatus calculates an analog gain value corresponding to the output volume value of the speaker by using Equation (1) described above. In step S4, the analogue gain value matching with the output volume value and the corresponding Can be stored.

7 is a diagram illustrating a process of processing a signal according to the embodiment of FIG.

Referring to FIG. 7, in the signal processing device storing the analog gain value and the digital gain value for processing the microphone input signal through the process of FIG. 6, when a guidance voice is output from the speaker (S5) The output volume value of the speaker is checked, and the analog gain value and the digital gain value corresponding to the confirmed output volume value are checked by referring to the stored information in step S4 (S6).

Then, the signal processor amplifies the analog signal inputted through the microphone according to the identified analog gain value (S7), and converts the amplified analog signal into the digital signal using the analog-digital converter (S8).

Thereafter, the signal processing apparatus removes the echo-type announcement sound from the digital signal converted in step S8 (S9), and amplifies the echo-canceled digital signal according to the digital gain value confirmed in step S6 (step S10) .

In step S10, the signal processing device amplifies the digital signal by restoring the digital signal by attenuating the analog gain value in step S3, and restores the digital signal as much as the analog signal is attenuated by using Equation (2) And amplifies the echo canceled digital signal according to the calculated digital gain value.

The amplified digital signal is used to recognize a voice (S11).

As described above, according to the present invention, when the guidance voice output from the speaker is input through the microphone in the form of echo, the gain of the microphone is appropriately adjusted to prevent clipping, thereby effectively removing echo mixed with the user's uttered voice, Is improved.

The signal processing method for pants-in speech recognition according to an embodiment of the present invention may be implemented in a form of a program readable by various computer means and recorded in a computer-readable recording medium.

It should be noted that the embodiments disclosed in the present specification and drawings are only illustrative of specific examples for the purpose of understanding, and are not intended to limit the scope of the present invention. It will be apparent to those skilled in the art that other modifications based on the technical idea of the present invention are possible in addition to the embodiments disclosed herein. Furthermore, although specific terms are used in this specification and the drawings, they are used in a generic sense only to facilitate the description of the invention and to facilitate understanding of the invention, and are not intended to limit the scope of the invention.

10: storage unit 20: first amplifier
30: analog-to-digital converter 40: echo canceller
50: second amplifier 60: speech recognizer
70: control unit 100: signal processing device

Claims (11)

A storage unit for storing analog gain value information of a microphone corresponding to an output volume value of a speaker for outputting a guidance voice;
A first amplifier for amplifying an analog signal inputted through the microphone according to an analog gain value stored in the storage unit;
An analog-to-digital converter for converting the analog signal amplified by the first amplifier into a digital signal;
An echo canceller for removing an echo corresponding to a guidance voice from the digital signal converted by the analog-to-digital converter;
A second amplifier for amplifying the echo-canceled digital signal; And
Wherein the analog gain value information of the microphone corresponding to the output volume value of the speaker is stored in the storage unit so that clipping of the output volume value of the speaker where clipping of the echo inputted through the microphone occurs, Stores the gain value in the storage unit and, when the output volume value of the speaker for outputting the guidance voice exceeds a preset value, confirms the corresponding attenuated analog gain value in the storage unit, A controller for controlling the first amplifier to amplify an analog signal input through the microphone and for controlling the second amplifier to recover a digital signal as the analog signal is attenuated;
And a voice recognition unit for recognizing the voice of the user. The method according to claim 1,
Wherein,
And a control unit for checking an output volume value at which clipping occurs in the echo input through the microphone when the guidance voice is output from the speaker and decreasing the analog gain value corresponding to the clipped output volume value to reduce the attenuated analog gain value And stores the clipped output volume value and the corresponding attenuated analog gain value in the storage unit. ≪ Desc / Clms Page number 20 > The method according to claim 1,
Wherein,
Wherein the analog gain value corresponding to the output volume value of the speaker is calculated using the following equation and stored in the storage unit.

If

If
AG: Analog gain value
v: Output volume value
T: Maximum volume value at which clipping does not occur
A: Undamped analog gain value
S: Output level of speaker The method of claim 3,
Wherein,
Calculating a digital gain value for recovering the digital signal as much as the analog signal is attenuated in the first amplifier using the following equation and controlling the amplifier to amplify the digital signal according to the calculated digital gain value Characterized in that the signal processing device for speech recognition of pants.

If

If
DG: Digital gain value
AG: Analog gain value
v: Output volume value
T: Maximum volume value at which clipping does not occur
A: Undamped analog gain value The method according to claim 1,
A voice recognizer for recognizing a speech voice inputted together with the echo through the microphone using the digital signal amplified by the second amplifier;
And a voice recognition unit for recognizing the voice of the user. The signal processing apparatus stores the analog gain value information of the microphone corresponding to the output volume value of the speaker for outputting the guidance voice so that clipping does not occur with respect to the output volume value of the speaker where clipping of the echo inputted through the microphone occurs Storing an attenuated analog gain value;
Confirming a corresponding attenuated analog gain value when the output volume value of the speaker from which the signal processing apparatus outputs the guidance voice exceeds a predetermined value;
Amplifying an analog signal input through the microphone according to an identified analog gain value;
Converting the amplified analog signal into a digital signal;
Removing the echo corresponding to the announcement voice in the converted digital signal; And
Amplifying a digital signal from which the echo is removed so that the signal processing device recovers the digital signal as much as the analog signal is attenuated;
/ RTI > The method of claim < RTI ID = 0.0 > 1, < / RTI > The method according to claim 6,
Wherein the storing step comprises:
Checking the output volume value at which clipping occurs in the echo input through the microphone when the signal processing apparatus outputs a guidance voice from the speaker;
Calculating an attenuated analog gain value such that clipping does not occur by reducing the analog gain value corresponding to the clipped output volume value; And
Storing the clipped output volume value and the corresponding attenuated analog gain value;
And a voice recognition unit for recognizing the voice of the user. The method according to claim 6,
Wherein the storing step comprises:
Wherein the signal processing apparatus calculates and stores an analog gain value corresponding to an output volume value of the speaker using the following equation.

If

If
AG: Analog gain value
v: Output volume value
T: Maximum volume value at which clipping does not occur
A: Undamped analog gain value
S: Output level of speaker 9. The method of claim 8,
Wherein the amplifying comprises:
Wherein the signal processor amplifies the echo-canceled digital signal by calculating a digital gain value for restoring the digital signal as much as the analog signal is attenuated using the following equation: Signal processing method.

If

If
DG: Digital gain value
AG: Analog gain value
v: Output volume value
T: Maximum volume value at which clipping does not occur
A: Undamped analog gain value The method according to claim 6,
Recognizing a speech uttered with the echo through the microphone using the amplified digital signal;
Further comprising the steps of: (a) inputting a speech signal to the speech recognition unit; A computer-readable recording medium having recorded thereon a program for performing a signal processing method for pants-in speech recognition according to any one of claims 6 to 10.

Images