KR20140015894A - Apparatus and method for estimating location of sound source - Google Patents

Abstract

The present invention relates to an apparatus and a method for estimating the location of a sound source and more specifically, to an apparatus and a method for estimating the location of a sound source which can accurately estimate the location of the sound source by dividing the sound source direction using a microphone structure, and calculating SRP-PHAT through a combination of microphone pairs capable of detecting a divided sound source direction area. The method for estimating the location of a sound source includes the steps of: acquiring sound from a plurality of microphones; calculating generalized cross correlation (GCC) of the sound acquired from combinations of two different microphones; dividing the sound source direction into a first, a second, a third and a fourth area by using the microphone structure, and calculating SRP-PHAT by combining the GCC corresponding to the first, the second, the third and the fourth area; and estimating a direction with maximum accumulation of SRP-PHAT as the location of the sound source by accumulating SRP-PHAT calculated based on time. [Reference numerals] (AA) Start; (BB) End; (S10) Acquire voice signal from a plurality of microphones; (S20) Calculate generalized cross correlation (GCC) of the sound acquired from combinations of two different microphones; (S30) Divide the sound source direction into a first, second, third and fourth area by using the microphone structure, and calculate SRP-PHAT by using GCC corresponding to the first, the second, the third and the fourth area; (S40) Estimate a direction with maximum accumulation of SRP-PHAT as the location of the sound source by accumulating SRP-PHAT calculated based on time

Description

Apparatus and method for estimating location of sound source}

The present invention relates to an apparatus and method for estimating a sound source, and more particularly, by dividing a sound source direction using a microphone structure, and calculating an SRP-PHAT through a combination of microphone pairs capable of detecting a divided sound source direction region. A sound source position estimation apparatus and method for accurately estimating the position of a sound source.

Sound source location technology uses sound sensors such as microphone arrays to locate sound sources and speakers, and is widely used in various engineering applications such as the 3D virtual stereo industry, speaker recognition of humanoid intelligent robots, and military remote monitoring systems. Applied, creating a lot of added value. Much effort is being put into research and development to locate microphones by arranging a microphone array by arranging a plurality of microphones in series or in parallel and analyzing the signals input to the microphones.

This sound source estimation algorithm compensates the input signal of each microphone according to the difference of sound wave arrival time due to the difference of sound wave propagation path from any sound source position to each microphone, and finds the angle where the signal power becomes the maximum and adds the sound source. We estimate by position.

Korean Unexamined Patent Publication No. 2009-0044314

The technical problem to be solved by the present invention is to segment the sound source direction using a microphone structure, and to accurately estimate the position of the sound source by calculating the SRP-PHAT through a combination of microphone pairs capable of detecting the divided sound source direction region. A sound source position estimation apparatus and method are provided.

According to an aspect of the present invention, there is provided a sound source position estimation method comprising: acquiring sound from a plurality of microphones; Calculating a cross correlation (GCC) for the sound obtained from two different microphone combinations; Calculating a SRP-PHAT by dividing a sound source direction into first to fourth regions using the microphone structure and combining the cross-correlation corresponding to the first to fourth regions; And accumulating the calculated SRP-PHAT for each predetermined time period, and estimating a direction in which the maximum SRP-PHAT is accumulated the most as a sound source location.

In the present invention, the plurality of microphones are arranged such that four microphones are orthogonal to each other and each pair of microphones faces each other.

In the present invention, the step of calculating the cross-correlation (GCC), the four mutually orthogonal microphones, each pair of microphones arranged to face each other, the mutual correlation of the four forward microphone pairs ( GCC) and two diagonal microphone pairs are calculated.

In the present invention, the step of calculating the SRP-PHAT, the step of dividing the sound source direction into the first to fourth regions by using a structure in which the four microphones are orthogonal to each other, and each pair of microphones face each other. ; And calculating an SRP-PHAT by combining a cross correlation (GCC) of one forward microphone pair and a cross correlation (GCC) of two diagonal microphone pairs corresponding to the respective areas. do.

In the present invention, the dividing may include dividing the 360-degree sound source direction into first to fourth regions by 90 degrees.

According to an aspect of the present invention, there is provided a sound source position estimation device comprising: a plurality of microphones for acquiring sound; Computing a correlation (GCC) for sound obtained from two different microphone combinations, dividing a sound source direction into first to fourth regions using the microphone structure, and corresponding to the first to fourth regions. A sound source search unit for combining the cross-correlation to calculate an SRP-PHAT; And a sound source position estimator for accumulating the calculated SRP-PHAT for each predetermined time period and estimating a direction in which the maximum SRP-PHAT is accumulated the most as a sound source position.

In the present invention, the plurality of microphones are arranged such that four microphones are orthogonal to each other and each pair of microphones faces each other.

In the present invention, the sound source search unit, a cross-correlation calculation unit for calculating a correlation (GCC) for the sound obtained from two different microphone combinations; An area dividing unit dividing a sound source direction into first to fourth areas by using the microphone structure; And an SRP-PHAT calculator configured to calculate an SRP-PHAT by combining the cross correlations corresponding to the first to fourth regions.

In the present invention, the cross-correlation calculation unit has a structure in which the four microphones are orthogonal to each other and each microphone pair is disposed to face each other, so that the four correlation microphones (GCC) and two diagonal lines are arranged. A cross correlation (GCC) of the directional microphone pairs is calculated.

In the present invention, the SRP-PHAT calculation unit, the cross-correlation (GCC) of one forward microphone pair corresponding to each of the first to the fourth region and the cross-correlation (GCC) of two diagonal microphone pairs It is characterized by calculating the SRP-PHAT by combining.

As described above, according to the present invention, the position of the sound source can be accurately estimated by dividing the sound source direction using a microphone structure and calculating the SRP-PHAT through a combination of microphone pairs capable of detecting the divided sound source direction region. have.

1 is a block diagram showing the configuration of a sound source position estimation apparatus according to an embodiment of the present invention.
FIG. 2 is a diagram for explaining sound source direction region division in FIG. 1.
FIG. 3 is a diagram illustrating SRP-PHAT calculation from any sound source direction region in FIG. 1.
4 is a diagram illustrating each sound source direction region of FIG. 1.
5 is a flowchart illustrating an operation of a sound source position estimation method according to an embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. Referring to the accompanying drawings, the same or corresponding components are denoted by the same reference numerals, do.

1 is a block diagram showing the configuration of a sound source position estimation apparatus according to an embodiment of the present invention.

Referring to FIG. 1, the sound source position estimating apparatus 10 includes a microphone 100, a sound source searching unit 200, and a sound source position estimating unit 300.

The microphone 100 obtains sound from at least one or more sound sources, and includes four microphones as a microphone 1 (M1), a microphone 2 (M2), a microphone 3 (M3), and a microphone 4 (M4). In the present embodiment, the microphone 100 is arranged such that four microphones are orthogonal to each other and each pair of microphones faces each other. That is, as shown in Fig. 2, four microphones M1-M4 are arranged at 45 degrees, 135 degrees, 225 degrees and 315 degrees counterclockwise, respectively. In addition, the microphone 1 (M1) and microphone 2 (M2) pairs are arranged to face each other with the microphone 3 (M3) and microphone 4 (M4) pairs, and the microphone 1 (M1) and microphone 4 (M4) pairs are microphone 2 ( M2) and microphone 3 (M3) pairs are arranged to face each other.

The sound source searching unit 200 calculates a cross correlation (GCC) for sounds obtained by two different microphone pairs, and divides a sound source direction into first to fourth regions using the microphone structure. The SRP-PHAT is calculated by combining the cross correlations corresponding to the first to fourth areas.

In the present embodiment, the sound source search unit 200 includes a GCC calculator 210, an area divider 220, and an SRP-PHAT calculator 230.

The GCC calculator 210 calculates a generalized cross correlation (GCC) for sounds obtained from two different microphone combinations, that is, microphone pairs. Here, the cross-correlation refers to the difference in arrival time of the sound received by each of the two microphones from the sound source. The GCC calculation equation is shown in Equation 1 below.

는 가중치 함수를 나타내고,

는 i번째 마이크를 통해 입력된 신호의 주파수 영역 값이며,

는 주파수 영역의 공액 복소수 값이다.In Equation 1,

Represents a weight function,

Is the frequency domain value of the signal input through the i-th microphone,

Is a conjugate complex value in the frequency domain.

A total of six GCCs can be calculated in the microphone structure of the present embodiment. That is, GCC of microphone 1 (M1) and microphone 2 (M2), GCC of microphone 2 (M2) and microphone 3 (M3), GCC of microphone 3 (M3) and microphone 4 (M4), microphone 4 (M4) and The GCC of the microphone 1 (M1), the GCC of the microphone 1 (M1) and the microphone 3 (M3), the GCC of the microphone 2 (M2) and the microphone 3 (M3) can be calculated. Here, the first four GCCs may be referred to as a forward GCC, and the rear two GCCs may be referred to as a diagonal GCC.

The region dividing unit 220 divides the sound source direction into the first to fourth regions Q1 to Q4 using the microphone M1 to M4 structure. Errors occur when calculating delay distances due to environmental disturbances or microphone structure. At this time, when mapping the correlation value of the data including the delay distance error to the angle using the cosine inverse function, the angle mapping error is small in the linearly approaching section.

In the present embodiment, the section in which such an angular mapping error occurs is divided into 90 degrees. When calculating the omni-directional direction (0 degrees-360 degrees) through one microphone pair combination without dividing the interval, an imaginary source is generated and sound source position performance is impaired. Therefore, in the present embodiment, as shown in FIG. 2, the first region Q1 is in the range of 315 degrees to 45 degrees, the second region Q2 is in the range of 45 degrees to 135 degrees and the third region Q3. ) Is divided into a 135 degree-225 degree interval, and the fourth region Q4 is divided into a 225 degree-315 degree interval.

The SRP-PHAT calculation unit 230 calculates the SRP-PHAT by combining the cross correlations GCC corresponding to the first to fourth regions Q1 to Q4, respectively. When calculating the SRP-PHAT, the SRP-PHAT is calculated by combining one forward GCC, two diagonal GCCs corresponding to each region, and three GCCs in total.

Here, the SRP calculation equation is as follows.

은 두 마이크 사이의 신호 도달 시간차를 나타낸다.

는 앞서 설명한 GCC 값이다. SRP 알고리즘은 해석상 GCC의 모든 마이크 조합을 이용한 시간차 추정 기법이라고 할 수 있다. Where q is the candidate position coordinate of the sound source, N is the number of microphones (4 in this embodiment),

Denotes the difference in signal arrival time between two microphones.

Is the GCC value described above. The SRP algorithm can be interpreted as a time difference estimation technique using all microphone combinations of GCC.

로 적용하면 입력 신호의 파워를 갖게 하고 위상의 차이를 이용하는 것으로, PHAT(phase transform)이라 한다. PHAT 가중치 기법은 반향에 강인하다고 알려져 있고, 이를 적용한 SPR 알고리즘을 SRP-PHAT 라 한다. Especially the weight function

In this case, the power of the input signal is used and the phase difference is used. This is called a PHAT (phase transform). The PHAT weighting technique is known to be robust to echo, and the SPR algorithm to which it is applied is called SRP-PHAT.

When calculating the SRP-PHAT in the first region Q1 315 to 45; FIG. 4A, the GCC of the forward microphone 1 M1 and the microphone 4 M4 and the diagonal microphone 1 corresponding to the first region Q1 are calculated. SRP-PHAT is calculated by combining the GCC of M1 and microphone 3 (M3), the GCC of microphone 2 (M2) and microphone 4 (M4), and a total of three GCCs. Here, the combination of GCC can be variously combined into summation, multiplexing, minimum, and the like.

When calculating the SRP-PHAT of the second area (Q2: 45 degrees to 135 degrees: FIG. 4B), as shown in FIG. 3, the forward microphone 1 (M1) and the microphone 2 (M2) corresponding to the second area (Q2). SRP-PHAT is calculated by combining the GCC of the first direction, the diagonal direction microphone 1 (M1) and the microphone 3 (M3) and the GCC of the microphone 2 (M2) and the microphone 4 (M4).

In calculating the SRP-PHAT in the third region Q3 (135 degrees to 225 degrees: FIG. 4C), the GCC of the forward microphone 2 (M2) and the microphone 3 (M3) and the diagonal microphone 1 corresponding to the third region Q3 are calculated. SRP-PHAT is calculated by combining the GCC of M1 and microphone 3 (M3) with the GCC of microphone 2 (M2) and microphone 4 (M4).

When calculating the SRP-PHAT in the fourth region (Q4: 225 degrees to 315 degrees: FIG. 4D), the GCC of the forward microphone 3 (M3) and the microphone 4 (M4) and the diagonal microphone 1 corresponding to the fourth region (Q4) are calculated. SRP-PHAT is calculated by combining the GCC of M1 and microphone 3 (M3) with the GCC of microphone 2 (M2) and microphone 4 (M4).

The sound source position estimator 300 accumulates the SRP-PHAT calculated for each predetermined time period and estimates the direction in which the maximum SRP-PHAT is accumulated the most as the sound source position. In the case of one sound source, the direction in which the maximum SRP-PHAT is accumulated the most is calculated. In the case of a plurality of sound sources, the direction in which the maximum SRP-PHAT is accumulated the most is calculated.

As described above, the position of the sound source can be accurately estimated by dividing the sound source direction using a microphone structure and calculating the SRP-PHAT through a combination of microphone pairs capable of detecting the divided sound source direction region.

5 is a flowchart illustrating an operation of a sound source position estimation method according to an embodiment of the present invention. In the following description, the description of the parts which are the same as those in the description of Figs. 1 to 4 will be omitted.

Referring to FIG. 5, the sound source position estimating apparatus 10 performs a step of obtaining sound from a plurality of microphones (S10). The microphone obtains sound from at least one sound source and includes four microphones. The structure of the microphones is arranged such that four microphones are orthogonal to each other, and each pair of microphones faces each other.

When sound is obtained from a plurality of microphones, the sound source position estimation apparatus 10 performs a step S20 of calculating a mutual correlation (GCC) with respect to sounds acquired by two different microphone pairs. A total of six GCCs can be calculated in the microphone structure of the present embodiment. That is, GCC of microphone 1 (M1) and microphone 2 (M2), GCC of microphone 2 (M2) and microphone 3 (M3), GCC of microphone 3 (M3) and microphone 4 (M4), microphone 4 (M4) and The GCC of the microphone 1 (M1), the GCC of the microphone 1 (M1) and the microphone 3 (M3), the GCC of the microphone 2 (M2) and the microphone 3 (M3) can be calculated. Here, the first four GCCs may be referred to as a forward GCC, and the rear two GCCs may be referred to as a diagonal GCC.

When the calculation of the cross correlation GCC is completed, the sound source position estimating apparatus 10 divides the sound source direction into the first to fourth regions Q1 to Q4 using the microphone M1 to M4 structure, and the first to the first region. In operation S30, the SRP-PHAT is calculated using the GCC corresponding to the fourth to fourth regions Q1 to Q4. The sound source position estimating apparatus 10 sets the section in which the angular mapping error is less than 90 degrees, the first region Q1 is 315 degrees to 45 degrees, and the second region Q2 is 45 degrees to 135 degrees. For example, the third region Q3 is divided into 135-225 degrees intervals, and the fourth region Q4 is divided into 225 degrees-315 degrees intervals. GCC of the forward microphone 1 (M1) and the microphone 4 (M4), the GCC of the diagonal microphone 1 (M1) and the microphone 3 (M3), and the microphone 2 (M2) with respect to the 1st area | region Q1: 315 degree-45. And combining the GCC of the microphone 4 (M4) and a total of three GCCs to calculate the SRP-PHAT, and the GCC of the forward microphone 1 (M1) and the microphone 2 (M2) with respect to the second region (Q2: 45-135 degrees). , Combining the GCC of the microphone 1 (M1) and the microphone 3 (M3) in the diagonal direction and the GCC of the microphone 2 (M2) and the microphone 4 (M4) to calculate the SRP-PHAT, and the third region (Q3: 135 degrees-225). FIG. 6) combines the GCC of forward microphone 2 (M2) and microphone 3 (M3), the GCC of diagonal microphone 1 (M1) and microphone 3 (M3), and the GCC of microphone 2 (M2) and microphone 4 (M4). Calculate the SRP-PHAT, and the GCC of the forward microphone 3 (M3) and the microphone 4 (M4), the diagonal microphone 1 (M1) and the microphone 3 (M3) for the fourth region (Q4: 225 degrees to 315 degrees). SRP-PHAT is calculated by combining the GCC and the GCC of the microphone 2 (M2) and the microphone 4 (M4).

When the calculation of the SRP-PHAT is completed, the sound source position estimating apparatus 10 accumulates the SRP-PHAT calculated for each predetermined time period and estimates the direction in which the largest SRP-PHAT is accumulated the most as the sound source position (S40). To perform.

The present invention has been described above with reference to preferred embodiments. It will be understood by those skilled in the art that the present invention may be embodied in various other forms without departing from the spirit or essential characteristics thereof. Therefore, the above-described embodiments should be considered in an illustrative rather than a restrictive sense. The scope of the present invention is shown not in the above description but in the claims, and all differences within the scope should be construed as being included in the present invention.

100: microphone 200: sound source search unit
210: GCC calculator 220: region divider
230: SRP-PHAT calculation unit 300: sound source position estimation unit

Claims (10)

Acquiring sound from the plurality of microphones;
Calculating a cross correlation (GCC) for the sound obtained from two different microphone combinations;
Calculating a SRP-PHAT by dividing a sound source direction into first to fourth regions using the microphone structure and combining the cross-correlation corresponding to the first to fourth regions; And
And accumulating the calculated SRP-PHAT for each predetermined time period, and estimating a direction in which the maximum SRP-PHAT is accumulated the most as a sound source location. The method of claim 1, wherein the plurality of microphones,
And four microphones orthogonal to each other, and each pair of microphones is disposed to face each other. The method of claim 1, wherein the calculating of the cross correlation (GCC) comprises:
Computing the correlation of the four forward microphone pairs (GCC) and the two diagonal microphone pairs (GCC) from a structure in which the four microphones are orthogonal to each other and each microphone pair is disposed to face each other. Sound source position estimation method, characterized in that. The method of claim 1, wherein the calculating of the SRP-PHAT comprises:
Dividing a sound source direction into first to fourth regions by using a structure in which four microphones are orthogonal to each other and each microphone pair is disposed to face each other; And
Calculating an SRP-PHAT by combining a cross correlation (GCC) of one forward microphone pair and a cross correlation (GCC) of two diagonal microphone pairs corresponding to the respective areas; Sound source position estimation method. The method of claim 4, wherein the dividing comprises:
A sound source position estimation method, comprising dividing a 360-degree sound source direction into first to fourth regions by 90 degrees. A plurality of microphones for acquiring sound;
Computing a correlation (GCC) for the sound obtained by the two different pairs of microphones, and divides the sound source direction into first to fourth regions using the microphone structure, the first to fourth regions A sound source search unit for combining the corresponding correlations to calculate an SRP-PHAT; And
And a sound source position estimating unit for accumulating the calculated SRP-PHAT for each predetermined time period and estimating a direction in which the maximum SRP-PHAT is accumulated the most as a sound source position. The method of claim 6, wherein the plurality of microphones,
And four microphones orthogonal to each other, and each pair of microphones is disposed to face each other. The method of claim 6, wherein the sound source search unit,
A cross-correlation calculation unit for calculating a cross-correlation (GCC) for sounds obtained from two different microphone combinations;
An area dividing unit dividing a sound source direction into first to fourth areas by using the microphone structure; And
And a SRP-PHAT calculation unit for combining the cross-correlation corresponding to the first to fourth areas to calculate an SRP-PHAT. The method of claim 8, wherein the cross-correlation calculation unit,
Computing the correlation of the four forward microphone pairs (GCC) and the two diagonal microphone pairs (GCC) from a structure in which the four microphones are orthogonal to each other and each microphone pair is disposed to face each other. Sound source position estimation device, characterized in that. The method of claim 8, wherein the SRP-PHAT calculation unit,
SRP-PHAT is calculated by combining the cross-correlation (GCC) of one forward microphone pair and the cross-correlation (GCC) of two diagonal microphone pairs corresponding to each of the first to fourth regions. Sound source position estimation device.

Images