KR20170107781A - Apparatus and method for discriminating sound source azimuth of mixed signal dynamically panned - Google Patents

Abstract

The present invention provides a device and a method for preventing sound quality degradation of separation sound source generated when separating the sound source with respect to a mixed stereo signal based on fixed panning and dynamic panning. Disclosed are a device and a method for identifying a sound source azimuth angle of a mixed signal dynamically panned. The method for identifying a sound source azimuth angle comprises the following steps: generating a frequency-azimuth plane of the mixed signal based on a real azimuth angle axis of the mixed signal; determining a cumulative energy for each azimuth angle by applying a time interval window to the frequency-azimuth plane; and identifying the azimuth angle having a peak value of the cumulative energy for each azimuth angle.

Description

[0001] APPARATUS AND METHOD FOR DISCRIMINATING SOUND SOURCE AZIMUTH OF MIXED SIGNAL DYNAMICALLY PANNED [0002]

The present invention relates to an apparatus and a method for identifying a source azimuth angle in a stereo audio signal in which a plurality of sound sources are dynamically panned.

A technique for separating a sound source from a stereo audio signal in which a plurality of sound sources are mixed can be utilized in various applications such as multi-channel upmixing based multi-channel audio service, object-based audio service, and sound source location tracking. Especially, considering the fact that most of the current audio content is produced by a stereo channel, it can have a great advantage in terms of technology utilization.

There is an Azimuth Discrimination and Resynthesis (ADRess) algorithm as a sound source separation technique for a conventional stereo channel. The ADRess algorithm is an algorithm that utilizes human auditory characteristics that recognize the position of a sound source based on the intensity difference (IID: Inter-aural Intensity Difference) between audio signals input to the left and right ears of a human.

However, the ADRess algorithm has a stable sound source separation performance using a cumulative energy of frequency-azimuth plane values for a mixed stereo signal with sound sources having a fixed panning index, while a sound source having a variable panning index is mixed In the case of the dynamic panning mixed sound source, the sound quality of the separated sound source is deteriorated.

In the case of dynamic panning where the position of a sound source changes in time, it is difficult to accurately identify the azimuth angle of the sound source unlike the fixed panning. In particular, the sound quality of the separated sound source is lowered because many errors are included in the silence interval.

Therefore, a method for preventing deterioration in sound quality of a separate sound source caused when a sound source is separated is also required for a dynamic panned mixed signal.

The present invention can provide an apparatus and method for preventing deterioration in sound quality of an isolated sound source caused when a sound source is separated from a mixed stereo signal based on fixed panning and dynamic panning.

A method for identifying a source azimuth according to an embodiment of the present invention includes: generating a frequency-azimuth plane of a mixed signal based on an actual azimuth axis of the mixed signal; Determining an accumulated energy for each azimuth angle by applying a time window to the frequency azimuth angle plane; And identifying an azimuth having a peak value of cumulative energy for each azimuth angle.

The generating the azimuth angle identification method according to an exemplary embodiment of the present invention includes generating a frequency azimuth plane of a mixed signal based on a signal intensity ratio between a left channel and a right channel of the mixed signal; Determining an actual azimuth using a relationship between a left channel and a right channel signal intensity ratio of the mixed signal and a relationship between the actual azimuth and the signal intensity ratio; And reconstructing a frequency azimuth plane of the mixed signal based on the determined actual azimuth angle axis.

The determining of the source azimuth identification method according to an embodiment of the present invention may include the step of sliding the time window on the reconstructed frequency azimuth plane and accumulating the energy of the reconstructed frequency azimuth plane along the azimuth axis have.

The time window of the method for identifying the source azimuth angle according to an embodiment of the present invention can be determined in consideration of an incorrect azimuth identification and a sensitivity to the azimuth identification that decrease as the width of the time window increases.

The step of identifying a source azimuth identification method according to an exemplary embodiment of the present invention may include a step of identifying an azimuth angle having a number of sound sources mixed with the mixed signal or a value equal to or greater than a reference value among cumulative energies for each azimuth angle included in each analysis frame of the mixed signal, Determining the cumulative energies as a peak value; And identifying the azimuth angle of the analysis frame based on the accumulated energy for each azimuth determined by the peak value.

A method of identifying a source azimuth angle according to an embodiment of the present invention includes: performing curve fitting at an azimuth angle identified in an analysis interval; And recalculating the azimuth angle of the analysis section using the curve fitting results.

The analysis interval of the sound source azimuth identification method according to an embodiment of the present invention may be set in consideration of at least one of a silent interval and a random azimuth error.

The azimuth identification device for a sound source according to an embodiment of the present invention includes a frequency azimuth plane generating unit for generating a frequency azimuth plane of a mixed signal based on an actual azimuth axis of the mixed signal; An accumulated energy determining unit for determining accumulated energy for each azimuth angle by applying a time window to the frequency azimuth angle plane; And an azimuth angle identification unit for identifying an azimuth angle having a peak value of cumulative energy for each azimuth angle.

The frequency azimuth plane generating unit of the source azimuth identifying apparatus according to an embodiment of the present invention generates a frequency azimuth plane of the mixed signal based on the signal intensity ratio between the left channel and the right channel of the mixed signal, The inter-channel signal intensity ratio and the relationship between the actual azimuth angle and the signal intensity ratio to determine the actual azimuth angle, and reconstruct the frequency azimuth plane of the mixed signal based on the determined actual azimuth angle axis.

The cumulative energy determiner of the source azimuth identification apparatus according to an embodiment of the present invention can slide the time window on the reconstructed frequency azimuth plane and accumulate the energy of the reconstructed frequency azimuth plane along the azimuth axis.

The time window of the source azimuth identification apparatus according to an embodiment of the present invention can be determined in consideration of an incorrect azimuth identification and a sensitivity to azimuth identification that decrease as the width of the time window increases.

The azimuth identification unit of the azimuth angle identification apparatus according to an embodiment of the present invention may be configured such that the azimuth angle identification unit of the azimuth angle identification unit of the azimuth angle identification unit of the azimuth angle identification unit of the azimuth angle discrimination unit includes a number of sound sources mixed with the mixed signal, The cumulative energy is determined as a peak value, and the azimuth of the analysis frame can be identified based on the accumulated energy for each azimuth determined by the peak value.

The apparatus for identifying a source azimuth according to an embodiment of the present invention includes: a curve fitting unit for performing curve fitting at an azimuth angle identified in an analysis section; And an azimuth calculator for recalculating an azimuth angle of the analysis section using a curve fitting result.

The analysis period of the source azimuth identification apparatus according to an embodiment of the present invention may be set in consideration of at least one of a silent period and a random azimuth error.

According to an embodiment of the present invention, a frequency azimuth plane of a mixed signal is generated based on a real azimuth axis of a mixed signal, and an accurate azimuth angle of a sound source is calculated based on cumulative energies of azimuths determined by applying a time window to a frequency azimuth plane It is possible to prevent deterioration in the sound quality of the separated sound source caused when the sound sources are separated from the mixed stereo signal based on the fixed panning and the dynamic panning.

FIG. 1 is a view showing a sound source azimuth identification apparatus according to an embodiment of the present invention.
2 is an example of a frequency azimuth plane, a time window, and an analysis interval according to an embodiment of the present invention.
3 is an example of a result of applying a time window in accordance with an embodiment of the present invention.
4 is an example of a result of applying curve fitting according to an embodiment of the present invention.
5 is a flowchart illustrating a method of identifying a source azimuth angle according to an embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. The method of identifying a source azimuth angle according to an embodiment of the present invention may be performed by a source azimuth identification device.

FIG. 1 is a view showing a sound source azimuth identification apparatus according to an embodiment of the present invention.

1, the sound source azimuth identification apparatus 100 includes a frequency azimuth plane generating unit 110, an accumulated energy determining unit 120, an azimuth identifying unit 130, an analysis interval setting unit 140, a curve fitting Unit 150 and an azimuth calculator 160. The azimuth estimator 160 may be a microprocessor, For example, the sound source azimuth identification device 100 may be a processor that performs a sound source azimuth identification method, or a storage medium that stores a sound source azimuth identification method. The frequency azimuth plane generating unit 110, the accumulated energy determining unit 120, the azimuth identifying unit 130, the analysis interval setting unit 140, the curve fitting unit 150, And may be each program module for performing the azimuth identification method.

The frequency azimuth plane generating unit 110 may generate a frequency-azimuth plane of the mixed signal based on the actual azimuth axis of the mixed signal. At this time, the mixed signal may be a mixed stereo signal based on fixed panning and dynamic panning.

First, the frequency azimuth plane generating unit 110 may generate a frequency azimuth plane of the mixed signal based on the signal intensity ratio between the left channel and the right channel of the mixed signal. At this time, the frequency azimuth plane generating unit 110 may generate a frequency azimuth plane plane for each analysis frame of the ADRess algorithm.

Specifically, the frequency azimuth plane generating unit 110 may apply STFT (Short-Time Fourier Transform) to each of the analysis frames of the mixed signal. The frequency azimuth plane generating unit 110 may generate the frequency azimuth plane of the (N + 1) x (beta + 1) array using the equation (1) in the STFT application result. In this case, N is the frequency resolution and? Can be the azimuth resolution. At this time, the azimuth resolution may be signal strength non-resolution.

일 수 있다. 또한, X₁(k,m)은 좌측 채널의 m번째 프레임에서의 k번째 주파수 성분이고, X₂(k,m)는 우측 채널의 m번째 프레임에서의 k번째 주파수 성분일 수 있다. 그리고, 좌채널과 우채널간 신호 강도비 g(i)는 0과 1사이의 값을 가질 수 있다. At this time, k is

Lt; / RTI > Further, X ₁ (k, m) may be the kth frequency component in the mth frame of the left channel and X ₂ (k, m) may be the kth frequency component in the mth frame of the right channel. Then, the signal intensity ratio g (i) between the left channel and the right channel may have a value between 0 and 1.

For example, the signal intensity ratio g (i) between the left channel and the right channel can be defined as Equation (2).

이며, i와 β는 정수일 수 있다. 이때, β 값이 증가할수록 방위각 식별의 정확도 및 계산량이 증가할 수 있다. 따라서, 방위각 식별의 정확도 및 계산량을 고려하여 β 값이 결정될 수 있다. 또한, ADRess 알고리즘을 이용한 주파수 방위각 평면에서 방위각 축은 신호 강도비 함수에서의 i 값일 수 있다. 즉, ADRess 알고리즘을 이용한 주파수 방위각 평면의 방위각은 i 값을 이용하여 추정한 방위각일 수 있다.In Equations (1) and (2), i is

, And i and? May be integers. At this time, as β value increases, accuracy and calculation amount of azimuth angle can be increased. Therefore, the beta value can be determined in consideration of the accuracy of azimuth angle identification and the calculation amount. Also, the azimuth axis in the frequency azimuth plane using the ADRess algorithm can be the i value in the signal intensity ratio function. That is, the azimuth angle of the frequency azimuth plane using the ADRess algorithm may be an azimuth estimated using the i value.

, A_Z이 최소가 되는 g(i)와 음원이 우측 채널에서 우세한 경우

, A_Z이 최소가 되는 g(i)를 찾을 수 있다. 이때, 주파수 방위각 평면 생성부(110)는 수학식 1을 재정의한 수학식 3을 이용하여 주파수 방위각 평면 전체를 구성할 수 있다. In addition, the frequency azimuth plane generating unit 110 generates a frequency azimuth plane when the sound source is dominant in the left channel

, G (i) where A _Z is the minimum and sound source is dominant in the right channel

, G (i) where A _Z is the minimum can be found. At this time, the frequency azimuth plane generating unit 110 can construct the entire frequency azimuth plane using Equation (3) redefining Equation (1).

In this case, max is A _z (kmi) maximum value on the azimuth angular line for the k-th frequency component, and min may be A _z (kmi) minimum value on the azimuth angular axis for the k-th frequency component. When the position of the sound source is on the left side, the azimuth angle is displayed on the right side. When the position of the sound source is positive, the azimuth angle is indicated by 90. When the position of the sound source is on the right side, As shown in FIG.

Next, the frequency azimuth plane generating unit 110 can determine the actual azimuth angle using the relationship between the left and right channel signal intensity ratios of the mixed signal and the actual azimuth angle and signal intensity ratio. At this time, the actual azimuth angle may be the accurate azimuth angle of the sound source.

At this time, the frequency azimuth plane generating unit 110 can determine an accurate source azimuth angle using the relation between the actual azimuth angle and the signal intensity derived based on the sinusoidal energy-preserving panning law.

For example, the frequency azimuth plane generating unit 110 may determine the azimuth azimuth (i) of the sound source using Equation (4).

이며, i와 β는 정수일 수 있다.At this time, i

, And i and? May be integers.

Finally, the frequency azimuth plane generating unit 110 can reconstruct the frequency azimuth plane of the mixed signal based on the determined actual azimuth angle axis.

At this time, the frequency azimuth plane generating unit 110 applies the signal intensity ratio g (i) determined using Equation (2) to Equation (4) to reconstruct the frequency-azimuth plane of each signal analysis frame based on the actual azimuth axis .

Specifically, the frequency azimuth plane generating unit 110 changes the i azimuth (i) representing the azimuth angle in the frequency azimuth plane plane generated using Equations (1) and (3) to the azimuth azimuth The plane can be reconstructed.

The cumulative energy determining unit 120 may determine the cumulative energies for each azimuth angle by applying a time window to the frequency azimuth plane plane generated by the frequency azimuth plane generating unit 110. [

Specifically, the cumulative energy determining unit 120 may accumulate the energy of the reconstructed frequency azimuth plane along the azimuth axis by sliding the time window on the reconstructed frequency azimuth plane.

은 수학식 5와 같이 정의될 수 있다.For example, when the cumulative energy determining unit 120 accumulates the planar value energy

Can be defined as shown in Equation (5).

At this time, W may be the width of the time window. As the width of the window between the time windows increases, it is possible to reduce the incorrect azimuth identification that occurs in random form. However, as the window width of the time window increases, the influence of the previous analysis frame increases and the sensitivity to the azimuth identification at the current time point may decrease. Therefore, the width of the time window can be determined in consideration of the reduction of the incorrect azimuth identification, which occurs in a random form, and the decrease in the sensitivity to azimuth identification at the present time.

The azimuth identification unit 130 can identify an azimuth angle having a peak value of cumulative energy for each azimuth angle.

들 중 혼합 신호에 혼합된 음원 개수, 또는 기준값 이상인 값을 가지는 방위각 별 누적 에너지를 피크값으로 결정할 수 있다. 그리고, 방위각 식별부(130)는 피크값으로 결정된 방위각 별 누적 에너지를 기초로 분석 프레임의 방위각을 식별할 수 있다. 구체적으로, 방위각 식별부(130)는 피크값으로 결정된 방위각 별 누적 에너지

의 i 값을 분석 프레임의 방위각으로 식별할 수 있다.At this time, the azimuth identification unit 130 compares the accumulated energy per azimuth angle included in each of the analysis frame m of the mixed signal

The cumulative energy of the azimuth angle having a value equal to or greater than the reference value can be determined as the peak value. The azimuth identification unit 130 can identify the azimuth angle of the analysis frame based on the cumulative energies of the azimuth angles determined as the peak values. Specifically, the azimuth identification unit 130 calculates the azimuth angle?

Can be identified by the azimuth of the analysis frame.

The analysis section setting section 140 can set an arbitrary analysis section. At this time, the analysis section may be set in consideration of at least one of a silent section and a random azimuth error.

The curve fitting performing unit 150 may perform curve fitting on the identified azimuth angle in the analysis period set by the analysis period setting unit 140. [

The azimuth calculator 160 may recalculate the azimuth angle of the analysis section using the curve fitting results output by the curve fitting performing unit 150. [

The sound source azimuth identification apparatus 100 generates a frequency azimuth plane of the mixed signal based on the actual azimuth axis of the mixed signal and identifies the accurate azimuth angle of the sound source based on the accumulated energy of each azimuth determined by applying the time window to the frequency azimuth plane Thus, it is possible to prevent deterioration of the sound quality of the separated sound source caused when the sound sources are separated from the mixed stereo signal based on the fixed panning and the dynamic panning.

2 is an example of a frequency azimuth plane, a time window, and an analysis interval according to an embodiment of the present invention.

The sound source azimuth identification device 100 determines the actual azimuth using the relationship between the left and right channel signal intensity ratios of the mixed signal and the actual azimuth angle and signal intensity ratio and sets the i value representing the azimuth angle in the frequency azimuth plane as the actual azimuth angle By changing to azimuth (i), the frequency azimuth plane 210 of the mixed signal can be reconstructed.

2, the sound source azimuth identification apparatus 100 can determine the planar value energies of the frequency azimuth plane 210 accumulated by the time window width by sliding the window 220 between the time periods.

Next, the sound source azimuth identification device 100 performs curve fitting at the identified azimuth angle in the analysis interval as shown in FIG. 2, and recalculates the azimuth angle of the analysis interval using the curve fitting result .

3 is an example of a result of applying a time window in accordance with an embodiment of the present invention.

In one embodiment, the dynamic panned mixed signal includes a first sound source having an azimuth angle varying from 20 to 70 degrees from the beginning to half of the entire interval, and an azimuth angle varying from 160 to 110 degrees from the beginning to half of the entire interval The second sound source having the second sound source.

The sound source azimuth identification apparatus 100 generates a frequency azimuth plane of the mixed signal based on the actual azimuth axis of the mixed signal and then applies the time window without applying the time window, The azimuth identification result 310 and the azimuth identification result 320 of the second sound source. At this time, as shown in FIG. 3, the azimuth identification result 310 of the first sound source and the azimuth identification result 320 of the second sound source indicate that the identification error with respect to the sound source azimuth angle in the silence interval of the first sound source and the second sound source becomes large A phenomenon can be displayed.

In addition, the sound source azimuth identification device 100 generates a frequency azimuth plane of the mixed signal based on the actual azimuth axis of the mixed signal, and then applies a rectangular window having a window width W of 5 as a time window It is possible to output the azimuth identification result 311 of the first sound source and the azimuth identification result 321 of the second sound source. At this time, the sound source azimuth identification apparatus 100 may apply a triangular window as a time window to reduce the influence of the frequency-azimuth plane value of the previous analysis frame.

Here, the azimuth identification result 311 of the first sound source and the azimuth identification result 321 of the second sound source are the same as the azimuth identification result 310 of the first sound source, which is a result of not applying the space time window, The azimuth identification error that occurs more randomly than the azimuth identification result 320 of the second sound source can be improved.

However, since the signal component X _i (k, m) does not exist in the silence period of the sound source, it is difficult to estimate g (i) for the sound source. Therefore, even if the time window is applied, the error identification error of the source azimuth in the silence period may not be improved.

Accordingly, the sound source azimuth identification device 100 can perform curve fitting as shown in FIG. 4 to re-identify the source azimuth angle.

4 is an example of a result of applying curve fitting according to an embodiment of the present invention.

The sound source azimuth identification device 100 sets an arbitrary analysis section in the mixed signal and adds curve fitting (first and second) to the azimuth identification result 311 of the first sound source and the azimuth identification result 321 of the second sound source curve fitting can be performed. Then, the sound source azimuth identification device 100 recalculates the azimuth angle value of the corresponding section by using the result of the curve fitting operation to re-identify the sound source azimuth angle so that the azimuth identification result 410 of the first sound source and the azimuth identification A result 420 can be output.

More specifically, the azimuth identification result 410 of the first sound source and the azimuth identification result 420 of the second sound source are inputted to the first-order linearly-polarized-angle identification result 311 and the second- This is the result of re-identifying the source azimuth angle by performing a linear polynomial curve fitting.

As shown in FIG. 4, the azimuth identification result 410 of the first sound source and the azimuth identification result 420 of the second sound source correspond to the azimuth identification result 311 of the first sound source and the azimuth identification result 321 of the second sound source, An identification error with respect to the sound source azimuth generated in the silent section can be improved.

5 is a flowchart illustrating a method of identifying a source azimuth angle according to an embodiment of the present invention.

In step 510, the frequency azimuth plane generator 110 may generate a frequency-azimuth plane of the mixed signal based on the actual azimuth axis of the mixed signal.

At this time, the frequency azimuth plane generating unit 110 may generate the frequency azimuth plane of the mixed signal based on the signal intensity ratio between the left channel and the right channel of the mixed signal. The frequency azimuth plane generating unit 110 can determine the actual azimuth using the relationship between the left and right channel signal intensity ratios of the mixed signal and the actual azimuth angle and signal intensity ratio. Finally, the frequency azimuth plane generating unit 110 can reconstruct the frequency azimuth plane of the mixed signal based on the determined actual azimuth angle axis.

In step 520, the cumulative energy determining unit 120 may determine a cumulative energy for each azimuth angle by applying a time window on the frequency azimuth plane in step 510. At this time, the cumulative energy determiner 120 may slice the time window on the reconstructed frequency azimuth plane, and accumulate the energy of the reconstructed frequency azimuth plane along the azimuth axis.

의 i 값을 분석 프레임의 방위각으로 식별할 수 있다.In step 530, the azimuth identification unit 130 may identify an azimuth angle having a peak value of the accumulated energy for each azimuth determined in step 520. At this time, the azimuth identification unit 130 may determine the cumulative energies of azimuth angles having the number of sound sources mixed in the mixed signal among accumulated energies for each azimuth angle included in each analysis frame of the mixed signal, or a value equal to or greater than a reference value as a peak value. Then, the azimuth identification unit 130 calculates the azimuth angle < RTI ID = 0.0 >

Can be identified by the azimuth of the analysis frame.

The curve fitting performing unit 150 may perform curve fitting at the azimuth angle identified in the analysis interval set by the analysis interval setting unit 140 among the azimuths identified at Step 530 at Step 540.

In step 550, the azimuth recalculator 160 may recalculate the azimuth angle of the analysis section using the curve fitting results in step 540.

The present invention creates a frequency azimuth plane of a mixed signal based on the actual azimuth axis of the mixed signal and identifies the exact azimuth of the source based on the accumulated energy for each azimuth determined by applying a time window to the frequency azimuth plane, It is possible to prevent degradation of the sound quality of the separated sound source caused when the sound source is separated from the mixed stereo signal based on the dynamic panning.

The method according to an embodiment may be implemented in the form of a program command that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The program instructions to be recorded on the medium may be those specially designed and configured for the embodiments or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Magneto-optical media, and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions include machine language code such as those produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

While the invention has been shown and described with reference to certain preferred embodiments thereof, it will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. This is possible.

Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined by the equivalents of the claims, as well as the claims.

110: frequency azimuth plane generating unit
120: cumulative energy determining unit
130: azimuth angle identification unit
140: Analysis section setting section
150: Curve fitting performing unit
160: azimuth calculation unit

Claims (14)

Generating a frequency-azimuth plane of the mixed signal based on the actual azimuth axis of the mixed signal;
Determining an accumulated energy for each azimuth angle by applying a time window to the frequency azimuth angle plane; And
A step of identifying an azimuth angle having a peak value of accumulated energy per azimuth
Wherein the azimuthal azimuth angle of the source azimuth is greater than the azimuth angle. The method according to claim 1,
Wherein the generating comprises:
Generating a frequency azimuth plane of the mixed signal based on the signal intensity ratio between the left channel and the right channel of the mixed signal;
Determining an actual azimuth using a relationship between a left channel and a right channel signal intensity ratio of the mixed signal and a relationship between the actual azimuth and the signal intensity ratio; And
Reconstructing the frequency azimuth plane of the mixed signal based on the determined actual azimuth axis
Wherein the azimuthal azimuth angle of the source azimuth is greater than the azimuth angle. The method according to claim 1,
Wherein the determining comprises:
Sliding the time window on the reconstructed frequency azimuth plane and accumulating the energy of the reconstructed frequency azimuth plane along the azimuth axis;
Wherein the azimuthal azimuth angle of the source azimuth is greater than the azimuth angle. The method according to claim 1,
The window of time-
Wherein the azimuth angle is determined in consideration of an incorrect azimuth identification and a sensitivity to azimuth identification that decrease as the width of the time window increases. The method according to claim 1,
Wherein the identifying comprises:
Determining cumulative energies of azimuth angles having a value equal to or greater than a reference number of sound sources mixed with the mixed signal among cumulative energies for each azimuth angle included in each of the analysis frames of the mixed signal as a peak value; And
Identifying an azimuth angle of the analysis frame based on accumulated energy for each azimuth determined by a peak value
Wherein the azimuthal azimuth angle of the source azimuth is greater than the azimuth angle. The method according to claim 1,
Performing curve fitting on the identified azimuth in any analysis interval; And
Calculating the azimuth angle of the analysis section using the curve fitting result
Further comprising the steps of: The method according to claim 6,
In the analysis section,
A silence interval, and a random azimuth error. A frequency azimuth plane generating unit for generating a frequency azimuth plane of the mixed signal based on an actual azimuth axis of the mixed signal;
An accumulated energy determining unit for determining accumulated energy for each azimuth angle by applying a time window to the frequency azimuth angle plane; And
An azimuth angle identification unit for identifying an azimuth angle having a peak value of accumulated energy for each azimuth angle
And an azimuth angle of the sound source. 9. The method of claim 8,
Wherein the frequency azimuth plane generating unit comprises:
The signal intensity ratio between the left channel and the right channel of the mixed signal and the relationship between the actual azimuth angle and the signal intensity ratio are used to generate the frequency azimuth plane of the mixed signal based on the signal intensity ratio between the left channel and the right channel of the mixed signal. Determines an azimuth angle, and reconstructs a frequency azimuth plane of the mixed signal based on the determined axis of the actual azimuth angle. 9. The method of claim 8,
Wherein the cumulative energy determining unit comprises:
And slides the time window on the reconstructed frequency azimuth plane to accumulate the energy of the reconstructed frequency azimuth plane along the azimuth axis. 9. The method of claim 8,
The window of time-
Wherein the azimuth angle is determined in consideration of incorrect azimuth identification and sensitivity to azimuth identification that decrease as the width of the time window increases. 9. The method of claim 8,
Wherein the azimuth-
A cumulative energy of each azimuth angle having a value equal to or greater than a reference value is determined as a peak value among the cumulative energies of the azimuth angle included in each of the analysis frames of the mixed signal and the accumulated energy per azimuth determined by the peak value And identifying the azimuth angle of the analysis frame based on the azimuth angle of the analysis frame. 9. The method of claim 8,
A curve fitting unit for performing curve fitting at the identified azimuth angle in an arbitrary analysis section; And
Calculating an azimuth angle of the analysis section using a curve fitting result,
Wherein the azimuth angle identification device further comprises: 14. The method of claim 13,
In the analysis section,
A silence interval, and a random azimuth error.

Images