KR101402287B1 - Apparatus for eliminating noise - Google Patents

Abstract

The noise canceller of the present invention includes a speaker for receiving a first signal, a microphone for outputting a second signal through a channel different from the channel of the speaker, and an echo canceller for eliminating interference between the first signal and the second signal Thus, various noises generated during voice communication can be reliably removed.

Description

[0001] APPARATUS FOR ELIMINATING NOISE [0002]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a noise canceling apparatus for eliminating noise generated during voice communication.

With the development of communication technology, it has become possible to transmit one voice to another. With this technology, a system for meeting with a remote party is provided.

In this way, noise removal is a very important problem in a system where voice is transmitted to another place of a remote place and the voice of a remote place is received and output.

Korean Patent Publication No. 0574666 discloses a technique of separately providing a noise input section in addition to a voice input section for collecting voice and canceling ambient noise using a noise input section.

However, there is a problem that noise can not be reliably removed due to a complicated and reliable configuration by providing a separate noise input section.

Korean Patent Registration No. 0574666

The present invention provides a noise canceling apparatus that reliably removes noise generated during voice communication.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory and are not intended to limit the invention to the precise forms disclosed. Other objects, which will be apparent to those skilled in the art, It will be possible.

The noise canceller of the present invention includes a speaker for receiving a first signal, a microphone for outputting a second signal through a channel different from the channel of the speaker, and an echo canceller for eliminating interference between the first signal and the second signal can do.

The noise canceller of the present invention can reliably remove noise generated during voice communication by eliminating interference between the signals of the speaker and the microphone using a separate channel.

In particular, the output sound of the speaker included in the output signal of the microphone can be reliably removed in an environment where the sound output from the speaker is micro-input.

In addition, the noise generated by the surrounding environment can be reliably removed.

1 is a schematic view showing a noise canceling apparatus of the present invention.
3 is another schematic diagram showing the operation of echo cancellation of the noise canceller of the present invention.
4 is a schematic diagram showing the operation of the echo cancellation of the present invention.
5 is a schematic view showing the voice enhancement unit of the present invention.
6 is a schematic diagram showing an actual configuration of a noise canceling apparatus according to an embodiment of the present invention.
7 is a schematic view showing the microphone of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. The sizes and shapes of the components shown in the drawings may be exaggerated for clarity and convenience. In addition, terms defined in consideration of the configuration and operation of the present invention may be changed according to the intention or custom of the user, the operator. Definitions of these terms should be based on the content of this specification.

1 is a schematic view showing a noise canceling apparatus of the present invention.

1 includes a speaker 110 to which a first signal is input, a microphone 130 for outputting a second signal through a channel different from that of the channel of the speaker 110, and an echo removal unit 150 ), And in the echo removal unit 150, the interference between the first signal and the second signal may be eliminated.

The speaker 110 is an element that receives the first signal, which is an electrical signal, and outputs it as a sound that can be perceived audibly. The voice at this time includes not only human voice but also various sounds and music that can be perceived audibly.

At this time, the first signal may be an electrical signal output from the external microphone 230 provided outside.

The microphone 130 is an element for receiving a voice and outputting a second signal as an electrical signal. At this time, the second signal may be provided to the external speaker 210 provided outside. The channel for providing the second signal to the external speaker 210 or the like by being connected to the microphone 130 may be a different channel from the channel used by the speaker 110. [ That is, since the speaker 110 and the microphone 130 included in the noise canceller of the present invention use separate channels, electrical signal interference between the speakers 110 and the microphone 130 is difficult to occur.

The noise eliminator performs communication between the speaker 110 at the first position and the external microphone 230 at the second position and performs communication between the microphone 130 at the first position and the external speaker 210 at the second position And a communication unit 190 for transmitting and receiving data. The communication unit 190 at this time can provide various types of wired / wireless communication interfaces supported by the communication network 300 and can communicate with the external terminal 200 at the second location. At this time, the second position may be a plurality of, and accordingly, the external microphone 230 and the external speaker 210 may also be plural.

The noise canceller may include an external terminal 200 having an external microphone 230 of the same channel as the speaker 110 and an external speaker 210 of a channel such as the microphone 130. At this time, the external speaker 210 converts the second signal into voice, and the external microphone 230 converts the mixed voice including the voice output from the external speaker 210 into the first signal and outputs the first signal .

The signal interference between the speaker 110 and the microphone 130, which are separated from each other by the channels, may be caused by auditory voice. Since the speaker 110 outputs voice and the microphone 130 receives voice, the voice output from the speaker 110 can be input to the microphone 130, Quot ;, and " noise "

For example, assume a system in which a first user at a first location and a second user at a second location meet.

Both the first user and the second user need to deliver their voice to the other party and listen to the other party's voice for the meeting. Accordingly, the speaker 110 and the microphone 130 are installed at the first position and the second position, respectively.

1 shows a state where a terminal equipped with the noise canceling apparatus of the present invention is disposed at a first position on the left side and an external terminal 200 having an external speaker 210 and an external microphone 230 disposed at a second position on the right side Lt; / RTI > For reference, a terminal equipped with the noise canceller of the present invention may be disposed at the second position.

When the first user at the first position speaks, it is input to the microphone 130 at the first position, and the input speech is converted to the second signal. The second signal is input to the external speaker 210 at the second position via the wired / wireless communication network 300. The external speaker 210 converts the inputted second signal into speech (second user's speech) and outputs it.

The voice output from the external speaker 210 is input to the external microphone 230 at the second position and converted into a first signal. The first signal is input to the speaker 110 at the first position via the communication network 300 and is outputted as a voice. Therefore, the first user becomes a situation in which he / she hears the words he / she has spoken. This situation can occur equally for a second user at the second location.

In order to solve this problem, it is necessary to separate the speaker 110 and the microphone 130 from each other so that the sound output from the speaker 110 is not input to the microphone 130. This method is applied to a system Can not.

Therefore, interference between the first signal and the second signal due to the auditory voice should be removed through other means.

The echo canceller 150 of the present invention can eliminate interference between the first signal and the second signal.

delete

delete

delete

delete

delete

delete

delete

delete

delete

The echo removal unit 150 may remove the other party's voice transmitted from the other party and introduced into the microphone 130, and may transmit only its own voice to the other party.

3 is another schematic diagram showing the operation of the echo removal unit 150 of the noise canceller of the present invention. In Fig. 3, a dotted line indicates a section in which auditory interference occurs.

When the speaker 110 converts the first signal into the first voice and outputs the converted voice, and when the microphone 130 converts the second voice including the first voice into the second voice and outputs the second voice, the echo removal unit 150 The first signal can be removed from the second signal.

The voice output from the speaker 110 is also input to the microphone 130. Accordingly, the second signal output from the microphone 130 includes the first signal input to the speaker 110. [ Thus, when the first signal is removed from the second signal, the output from the speaker 110 is removed. At this time, since the first signal and the second signal are signals communicated in the same terminal equipped with the noise canceller, it is not necessary to grasp the characteristic information of the encoder and decoder of the external terminal 200.

However, the sound output from the speaker 110 must be considered to be deteriorated by the surrounding environment during the input of the sound into the microphone 130. In other words, the second signal is converted into an electrical signal by converting the second voice including the first voice inputted into the microphone 130 into a state that the first signal is outputted from the speaker 110 to which the first signal is input and is changed by the surrounding environment, .

As a result, if the first signal is directly removed from the second signal, it may be difficult to obtain a desired result. For example, part of the voice of the first user may also be removed or part of the voice of the second user may not be removed.

In order to prevent such a phenomenon, the echo removal unit 150 may include an adaptive digital filter (ADF) that minimizes the power of a signal obtained by subtracting the first signal from the second signal.

FIG. 4 is a schematic diagram showing the operation of the echo removal unit 150 of the present invention. The function of the adaptive filter included in the echo removal unit 150 is disclosed.

When the first signal received from the external terminal 200 is x (n), the component of the first signal included in the second signal is y (n), not x (n) The interval ⓒ may be a period from when the first voice is output from the speaker 110 to when it is input to the microphone 130. It is not a concept of the distance but a noise such as a noise or a wall surface that can change the first voice It can be an environmental concept that includes.

를 생성하고 제2 신호에서

을 제거하는 방안을 취한다. 이때,

의 생성에 적응 필터를 이용할 수 있다.From the viewpoint of the echo removal section 150, it is difficult to grasp directly the element of y (n) because it is difficult to grasp the element of the section ⓒ which changes x (n) to y (n). Therefore, the echo removal unit 150 bypasses x (n) as close as possible to y (n)

Lt; RTI ID = 0.0 >

. At this time,

An adaptive filter can be used.

을 생성하기 위해 적응 필터의 계수를 적절하게 조절해야 한다. 이때, 에코 제거부(150)는 제2 신호에서 제1 신호를 차감한 신호의 전력을 최소화시키는 적응 필터의 계수 w_i(n)을 다음의 수학식 1로부터 산출할 수 있다. 수학식 1로 산출된 적응 필터의 계수 w_i(n)는 주파수 영역에서 적용될 수 있다.As close to y (n) as possible

The coefficients of the adaptive filter must be adjusted appropriately. At this time, the echo removal unit 150 may calculate the coefficient w _i (n) of the adaptive filter that minimizes the power of the signal obtained by subtracting the first signal from the second signal from Equation (1). The coefficient w _i (n) of the adaptive filter calculated by Equation (1) can be applied in the frequency domain.

Here, L is the number of samples of one block to be processed (a signal processing unit of the echo removal unit 150, for example, 256 or 512 data samples)

이며,

Lt;

로, 에코 제거부(150)를 거친 후의 제1 신호이고,

Is the first signal after passing through the echo removing unit 150,

x (n-1-i) is the i-th first signal before passing through the echo removal unit 150,

y (n-1) is a component of the first signal included in the second signal before passing through the echo removal unit 150,

P _x (n-1) is the smoothed power of x (n-1)

이며,

Lt;

lambda is a smoothing constant between 0 and 1,

μ is the adaptation coefficient

δ is a constant for making the denominator a positive, nonzero number.

When a signal having a large number of impulse components is outputted as it is, the user can not recognize the voice. Thus, the impulse signal is smoothly smoothed so that the user can recognize the voice. In Equation (1), the signal is smoothed through P _x (n-1).

The echo removal unit 150 may use a normalized complex least squares average method in the frequency domain. The normalized complex least squares average method may be a so-called LMS (Least Mean Square) algorithm. If w _i is a complex coefficient of the ADF may be a Complex LMS la, and a formula is defined to use the μ value between 0 and 1 normalized (Normalized) complex least mean square method.

In addition, it is determined whether the two-way communication state is established through the correlation measurement method of the first signal and the second signal, and the first signal component included in the second signal is reliably removed by adjusting the gain of the output signal .

The adaptation coefficient of Equation (1) can be determined by estimating a double talk interval in which the first user and the second user are simultaneously talking and an interval in which the other user (second user) is talking. That is, the adaptive coefficient μ can be used as a larger value in a section where only the other party talks. In a section where both parties talk at the same time, the data quality of the filter output can be obtained with a more effective value while using μ as a small value or setting it to 0 (zero) to ensure stability so that the adaptive filter does not diverge.

delete

delete

The echo canceller 150 described with reference to FIGS. 1 and 3 can reliably prevent an error generated when the voice output from the speaker 110 is input to the microphone 130 again.

On the other hand, it is advantageous to emphasize only the voice of the user by removing various ambient noises flowing into the microphone 130 in the interval c.

To this end, the noise canceller may include a voice enhancement unit 170.

As shown in FIG. 1, the voice enhancer 170 may be disposed in the channel of the microphone 130 at the output end of the echo canceller 150.

The speech enhancement unit 170 may estimate a noise component included in the second signal from at least one of the change amount of the frequency domain of the second signal and the change amount of the time domain and may attenuate the estimated noise component. For reference, the signals described below are assumed to be the signals passed through the echo removal unit 150. [

The second signal includes various ambient noise as well as the user's voice. At this time, the noise component can be reliably removed by dividing the noise component and the target signal corresponding to the user's voice.

In order to distinguish the noise component from the target signal, the speech enhancement unit 170 may estimate the noise component as follows.

Generally, a short-channel noise cancellation system is performed in the frequency domain and can estimate the size or noise of speech by determining attenuation or gain information of each frequency component. For example, when a mixture of voice and noise is input, the noise can be estimated using a characteristic in which the noise is relatively less varied than the voice in the short range.

5 is a schematic view showing the voice enhancement unit 170 of the present invention.

The speech enhancer 170 extracts the noise D (k, l) from the size information of the frequency component Y (k, l) of the input signal y (t) to which the noise d (t) Is estimated. The speech enhancement unit 170 estimates the gain G (k, l) using the estimated power spectrum and performs inverse fast Fourier transform (FFT) after noise spectral subtraction on the magnitude signal spectrum of the input signal Y Transforms can be used to synthesize speech.

If the processing section is estimated as a noise section including noise, the voice enhancer 170 reduces the size of the input signal in the gain controller. If the gain controller is not used, the voice enhancer 170 subtracts the frequency component of the noise And outputs the remaining component.

The variance of the frequency axis and the variation of the time axis among the statistical characteristics of the output signal of the microphone 130 can be calculated. In addition, a threshold value (set value) can be set by experimentally investigating the change amount of the noise interval and the target signal interval. By comparing the threshold at this time with the calculated variation, the noise signal can be distinguished from the target signal, and the noise signal can be estimated accordingly.

, 처리 중인 프레임의 전체 파워를 P_if라 할 때, 주파수 영역에서 정규화된 변화량(주파수 영역에서의 분산), 즉 주파수 편평도는 다음의 수학식 2와 같다.The power of each frequency component in the frequency domain is _{expressed as} the average of P _f , P _f

, And the total power of the frame under processing is P _if , the normalized amount of change (dispersion in the frequency domain) in the frequency domain, that is, the frequency flatness is expressed by Equation 2 below.

이다.At this time,

to be.

, 해당 프레임의 전체 파워를 P_it라 할때, 시간 영역에서 정규화된 변화량(시간 영역에서의 분산)은 수학식 3과 같다.The power of one frame in the time domain is the average of P _t , P _t

, And the total power of the frame is P _it , the normalized change amount in the time domain (variance in time domain) is expressed by Equation (3).

이다.At this time,

to be.

The target signal can be estimated when the values of the following Equation 4 calculated by IIR (Infinite Impulse Response) averaging of Equations (2) and (3) are greater than the experimentally obtained threshold value.

At this time,? Is an IIR smoothing coefficient,

P _{i, av} is the IIR average power.

If the target signal is estimated, a noise signal can be estimated relatively and an environment in which the estimated noise signal can be removed is provided.

As another parameter that can estimate the target signal, the IIR average for the power of the input signal can be calculated, and then the frame power can be averaged over a certain multiple compared with the current frame power. For a signal over a certain multiple, it is not included in the average calculation, so that it can be estimated as a target signal when a large signal is suddenly input continuously.

The IIR average power of the current frame can be calculated by Equation (5).

Where? Is an IIR coefficient between 0 and 1,

P _{i-1, av} is the IIR average power of the previous frame,

P _i is the power of the current frame.

If P _{i-1, av} corresponds to a frame less than a certain multiple of P _i , the IIR average is recalculated as shown in Equation (6).

Where alpha is an IIR smoothing coefficient between 0 and 1,

P _{i, Lav} is the long term IIR average power to the current frame,

P _i-1 is the power of the previous frame.

P _{i, Lav} calculated in Equation (6) do not apply an input signal with a suddenly large amount of change _, so that P _i, L _av represents the frame power level of the noise signal as a whole.

Therefore, when the power of the current frame of the input signal is larger than the set multiple of P _{i and Lav} as shown in Equation (7) _, the target signal can be estimated and protected.

At this time, P _i is the power of the current frame,

c is a constant experimentally obtained with a set multiple that can be estimated as a target signal.

In summary, the voice enhancement unit 170 can calculate an IIR (Infinite Impulse Response) average for the power of the second signal. At this time, the speech enhancer 170 may exclude the target period having the power equal to or greater than the set multiple of the IIR average in the second signal in the IIR average calculation process.

Also, the speech enhancement unit 170 may estimate the IIR average as the power of the noise component included in the second signal, and may attenuate the noise component.

The voice enhancement unit 170 of FIG. 5 calculates the frame power P _n of the noise period by using Equation (4) and Equation (7) The IIR average signal-to-noise ratio between the frame power P _n and the frame power P _i of the input signal of the microphone 130 can be calculated as Equation (8).

In this case, P _i-1 is the power of the previous frame,

beta is an IIR smoothing coefficient between 0 and 1,

abs is the absolute value operator.

The voice enhancement unit 170 of the present invention may use a Wiener filter to remove the estimated noise signal. Applying Equation (8) after application of the Witer filter, the weight value multiplied by the frequency component of the input signal can be expressed as Equation (9).

The gain multiplied by the frequency component of the input signal in Fig. 5 due to the deformation of the Wiener filter can be expressed by the following equation (10).

In this case, A _L is the attenuation coefficient of the noise component, and the larger the noise component is, the greater the attenuation is.

In order to reduce the abrupt change in gain, the IIR average for the gain is calculated as shown in Equation (11), and the frequency components of the input signal of the microphone 130 are multiplied.

Where λ is a constant between 0 and 1 determined experimentally.

6 is a schematic diagram showing an actual configuration of a noise canceling apparatus according to an embodiment of the present invention.

An ARM 920T processor or a general purpose processor and a real-time processing board based on a Philips CODEC UDA1341TS or AD (analog-digital) converter were developed based on the block diagram shown in FIG.

The input signals coming from the four microphones 130 are amplified by a preamplifier 710 including a first-order low-pass circuit with a cut-off frequency of 17 KHz, UDA1341TS stereo codec 720 to a digital value. The echo removal unit 150 configured by the program operates in 920T (740) to remove the echo component included in the input signal and the static noise of the environment in real time, and frequently used parameter values are automatically stored in the EEPROM 790, We designed the hardware for user convenience to maintain the optimum state adapted to the user. A button 730 and a connection unit 750 for connecting and disconnecting data to and from a PC or other device are provided so that the function can be changed in detail according to the environment of the user. 760) was used. In order to protect the copyright of the software, a security chip 770 for preventing copying is mounted. In order to facilitate the use state and control of the equipment, an LED for indicating the operation state of the program and whether or not the button is used is mounted on the front panel unit 780 Respectively.

The output sound of the speaker 110 and ambient noise input to the microphone 130 can be reduced through the structure of the microphone 130. [

7 is a schematic diagram showing the microphone 130 of the present invention. The left side of FIG. 7 shows the arrangement of the entire microphone 130, and the right side shows a cross section of one microphone 130 module.

7, the noise eliminator of the present invention includes a microphone 130, a microphone 130, and a microphone 130. The microphone 130 is installed in the housing 130, (120).

The housing 120 may take the form of a funnel as a whole. The housing 120 having such a shape collects the voice generated from the target sound source and can restrict the noise generated from the position distant from the target sound source.

A first groove 121 for absorbing sound of the first band is formed in a portion where the cross-sectional area of the housing 120 increases, a first groove 121 is formed in the first groove 121, A second groove 122 of a smaller size can be formed.

If a large number of grooves are formed acoustically in the space where the sound waves propagate, the sound waves are reflected at various angles in the grooves, so that the amount of propagation in the traveling direction decreases. A plurality of grooves can be formed on the inner surface of the housing 120 using this principle. At this time, the groove may include a first groove 121 and a second groove 122. Depending on the size of the groove, the frequency of the voice being canceled or absorbed in the groove is different. The larger the groove, the lower the frequency band sound is, and the smaller the groove, the higher frequency band sound is attenuated. Accordingly, the inner surface of the housing 120 may have a plurality of grooves of different sizes.

In order to improve space utilization and to provide a beautiful appearance, the second groove 122 of a small size can be formed in the first groove 121 of a larger size. In Fig. 7, the dotted line represents the first groove 121, and the small-sized groove formed about the dotted line represents the second groove 122. In Fig.

The first groove 121 absorbs the voice of the bass sound relatively as compared with the second groove 122 and the second groove 122 absorbs the relatively high voice sound as compared with the first groove 121. In addition, since the second groove 122 is formed on the inner surface of the first groove 121, as compared with the case where the first groove 121 and the second groove 122 are formed on the same surface in the housing 120, The first groove 121 and the second groove 122 can be formed. Further, since a simple image can be provided when viewed from the direction in which the sound source is present, a beautiful appearance can be provided.

A plurality of microphone 130 modules including the microphone 130 and the housing 120 may be provided. For example, in FIG. 7, four microphone modules 130 are formed in a row.

The shape of the housing 120 and the first groove 121 and the second groove 122 can reliably prevent the output sound or noise of the speaker 110 from being introduced even in the step of acquiring the second signal.

The above noise canceller may be applied to a conference system capable of bidirectional communication between the first position and the second position, a television system provided with a conference function, and the like.

While the invention has been shown and described with reference to certain preferred embodiments thereof, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention as defined by the appended claims. Accordingly, the true scope of the present invention should be determined by the following claims.

100 ... Noise canceling device 110 ... Speaker
120 ... housing 121 ... first groove
122 ... second groove 130 ... microphone
150 ... echo removing unit 170 ... sound enhancing unit
190 ... communication unit 200 ... external terminal
210 ... External speaker 230 ... External microphone
300 ... network

Claims (12)

A speaker to which a first signal is input;
A microphone for outputting a second signal through a channel different from the channel of the speaker;
An echo canceller for eliminating interference between the first signal and the second signal; And
And a housing in which the microphone is installed and the cross-sectional area gradually increases from a face on which the microphone is installed toward a direction toward a sound source to be targeted by the microphone,
Wherein a first groove for absorbing sound of a first band is formed at a portion where the cross-sectional area of the housing increases, a second groove which absorbs sound of a second band and is smaller than the first groove, Gt;
delete delete delete delete The method according to claim 1,
Wherein the second signal is a signal obtained by converting the second voice including the first voice, which has been output from the speaker to which the first signal is input, and converted by the surrounding environment into an electrical signal,
Wherein the echo canceller includes an adaptive filter that minimizes power of a signal obtained by subtracting the first signal from the second signal.
The method according to claim 6,
Wherein the echo canceling unit calculates the coefficient w _i (n) of the adaptive filter that minimizes the power of the signal obtained by subtracting the first signal from the second signal by the following equation.

Where L is the number of samples of a block being processed,

Lt;

, Which is a first signal after passing through the echo canceller,
x (n-1-i) is the i-th first signal before passing through the echo canceller,
y (n-1) is a component of the first signal included in the second signal before passing through the echo canceller,
P _x (n-1) is the smoothed power of x (n-1)

Lt;
lambda is a smoothing constant between 0 and 1,
μ is the adaptation coefficient
δ is a constant for making the denominator a positive, nonzero number.
delete delete delete delete delete

Images