KR101815584B1 - Compensation method for bias errors in 3D intensity probe used for the source localization in precision - Google Patents

Abstract

Provided is a method for compensating a bias error of a three-dimensional sound intensity probe device for precisely estimating a location of a sound source. According to an embodiment of the present invention, a method for estimating a location of a sound source generates and compensates mutual power spectral density functions between signals measured in a plurality of microphones, and then combines the mutual power spectral density functions to calculate a sound intensity vector. Therefore, a bias error on a three-dimensional sound intensity vector measured in a sound intensity probe composed of a micro microphone array is compensated, thereby precisely estimating a location of a sound source.

Description

BACKGROUND OF THE INVENTION Field of the Invention The present invention relates to a method of compensating a deflection error of a three-dimensional acoustic intensity sensing device for precisely estimating a position of a sound source,

BACKGROUND OF THE INVENTION Field of the Invention [0002] The present invention relates to a sound source position estimation technique, and more particularly, to a three-dimensional acoustic intensity sensing device.

The position estimation system according to the related art uses a microphone with high performance or increases the number of microphones to secure estimation performance. Currently widely used commercial three-dimensional intensity probes have six microphones orthogonal to each other. Based on the arithmetic mean of the measured sound pressure. However, due to the influence of the structures for supporting each microphone, there is room for carrying the error in the intensity measurement.

In a previous study in which the position of a sound source is estimated by using an acoustic intensity, a plurality of arrays including one or more microphones, including a head portion corresponding to a human face of a robot, There is also a sound source position estimation system for finding an acoustic intensity vector that is spatially incident from the sound source position estimation system.

In addition, there are some cases in which two bidirectional transducers are used to improve the position estimation performance, but this shows a position estimation error in a narrow frequency band or at a specific frequency.

Although two acoustic intensities are measured and the position of the sound source is estimated with the average value, it is possible to increase the accuracy of estimating the position of the sound source by using a plurality of acoustic intensities. However, there is a problem that a bias error appears.

In order to realize a small sound source localization system, the distance between the microphones must be narrowed, so that the frequency band to be measured can be narrowed or the resolution of the position estimation can be lowered.

When the beamforming method is applied, the number of microphones is basically required and the physical size of the system becomes large. Therefore, when the TDOA method is applied, the resolution is determined according to the interval between the microphones As the distance between the microphones decreases, the resolution becomes worse. Therefore, there is a limit in that the sampling frequency must be increased for accurate sound source position estimation.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a method and apparatus for accurately estimating the position of a sound source using a three dimensional acoustic intensity vector measured by an acoustic intensity transducer composed of a micro- And an error compensation method.

According to an aspect of the present invention, there is provided a method for estimating a sound source position, the method comprising: generating mutual power spectral density functions between signals measured in a plurality of microphones; Compensating the generated mutual power spectral density functions; And combining the compensated mutual power spectral density functions to yield an acoustic intensity vector.

And, the compensating step may linearize the phase functions of the generated mutual power spectral density functions.

Further, the phase function? Is expressed by the following equation,

? = 2? t ₀ f,

Where < RTI ID = _0.0 > # ₀ < / RTI > may be a time delay depending on the distance between the microphones.

The compensating step may perform low pass filtering on the generated mutual power spectral density functions.

Also, the cut-off frequency time by the low-pass filtering may be determined according to the distance between the microphones.

The low-pass filtering may be FIR filtering or IIR filtering.

In addition, the components of the acoustic intensity vector may be computed from the imaginary parts of the mutual power spectral density functions.

The method of estimating a sound source position according to an embodiment of the present invention may further include correcting the calculated sound intensity vector according to the frequency of the sound.

In addition, the correcting step may compensate for a deviation error according to an incidence direction of the sound.

The calculated acoustic intensity vector can be corrected by referring to a map defining at least one of a direction coefficient, an azimuth angle and an altitude angle according to the frequency of the sound.

In addition, the plurality of microphones may be four microphones arranged at vertices of a regular tetrahedron.

The compensating step may compensate the generated mutual power spectral density functions to eliminate a bias error caused by a plurality of microphones.

In addition, the deviation error may be caused by an interference between the incident wave and the reflected wave from the sound source.

According to another aspect of the present invention, there is provided an apparatus for estimating a sound source position, comprising: a generator for generating mutual power spectral density functions between signals measured in a plurality of microphones; A compensation unit for compensating the generated mutual power spectral density functions; And a calculation unit for calculating the acoustic intensity vector by combining the compensated mutual power spectral density functions.

As described above, according to the embodiments of the present invention, it is possible to precisely estimate the position of a sound source by compensating a deviation error of a three-dimensional sound intensity vector measured by an acoustic intensity transducer composed of a micro-array of microphones .

In addition, according to the embodiments of the present invention, it is possible to apply to a miniaturized robot, an ICT device, and military equipment as a high-precision, compact, and economical sound source location tracking system.

1 shows a three-dimensional acoustic intensity transducer,
FIG. 2 is a block diagram of a three-dimensional acoustic intensity sensing device according to an embodiment of the present invention,
FIG. 3 is a map showing a direction coefficient, an azimuth angle, an error range and a standard deviation of an altitude angle according to the change of kd,
4 is a graph showing the results of an experiment on an impact sound in an actual anechoic room condition.

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

1. Three-dimensional acoustic intensity probe

1 is a view showing a three-dimensional acoustic intensity transducer. As shown in Fig. 1, a three-dimensional acoustic intensity probe includes four microphones arranged at vertices of a regular tetrahedron.

The three-dimensional acoustic intensity transducer shown uses an acoustic intensity as a sound source position estimation method suitable for miniaturization in order to estimate an unknown sound source by using a micro array of three-dimensionally arranged microphones, and a deviation error occurs depending on the frequency .

Therefore, in the embodiment of the present invention, a method of precisely estimating a sound source position by grasping a cause of a bias error in performing a sound source position estimation using a three-dimensional acoustic intensity transducer and compensating for a bias error based on the cause do.

Specifically, it is clarified that the cause of the deviation error, which largely varies according to the frequency, is the phenomenon caused by the interference between the incident wave and the reflected wave, and the phase difference of the cross power spectral density function of the two waves is linearized or filtered to compensate We suggest a method.

Furthermore, a directional distribution DB is constructed by analyzing the directivity between a remote source and a transducer at an arbitrary position, and a method of compensating a deviation error according to the incidence direction is also presented.

FIG. 2 is a block diagram of a three-dimensional acoustic intensity sensing apparatus according to an embodiment of the present invention. 2, the three-dimensional acoustic intensiveness probe according to an embodiment of the present invention includes microphones 111 to 114, a cross power spectral density (CPSD) function generation unit 120, a CPSD function compensation unit 130, an acoustic intensity calculation unit 140, and an acoustic intensity correction unit 150.

The microphones 111 to 114 refer to four microphones arranged at vertices of a regular tetrahedron, as shown in FIG.

The CPSD function generation unit 120 generates CPSD functions between the sound signals generated by the microphones 111 to 114. [ The CPSD function compensating unit 130 compensates for the bias error with respect to the CPSD function generated by the CPSD function generating unit 120. [

The deflection error compensation is achieved by linearization or filtering of the CPSD function, which will be described in detail later.

The acoustic intensity calculation unit 140 combines the compensated CPSD functions in the CPSD function compensation unit 130 to calculate an acoustic intensity vector. The components of the acoustic intensity vector can be calculated from the imaginary parts of the CPSD functions.

The acoustic intensity correction unit 150 corrects the acoustic intensity vector calculated by the acoustic intensity calculation unit 140 according to the frequency of the sound. The position of the sound source can be estimated from the corrected sound intensity vector. The acoustic intensity vector correction method will be described later in detail.

2. Deflection Error of 3-D Acoustic Intensity Sensor

The three-dimensional acoustic intensity intensifying apparatus shown in FIG. 2 is a system in which the CPSDs between the microphones 111 to 114 are calculated in order to calculate a three-dimensional acoustic intensity vector at the acoustic center (the same as the geometric center of the probe) Function.

As described above, in this process, a deviation error occurs due to a spectral fluctuation having a specific periodicity different from a random error of a frequency band.

In the components constituting the imaginary part of the CPSD functions between the acoustic signals measured in the different microphones 111 to 114, the sine function based on the time component with high correlation other than the time delay due to the interval between the microphones causes the frequency fluctuation phenomenon . In other words, it is a phenomenon caused by the interference between the incident wave and the reflected wave.

The acoustic intensity vector component consists of the imaginary part of the CPSD function of the signals measured at the microphones 111-114. The CPSD function by the signals p ₁ and p ₂ , which are propagated by a single-pole source from two different microphones, can be expressed as a series of impulse functions as follows:

(1)

(One)

Where τ ₀ is the time delay of the incoming signal to each microphone.

이라고 가정한다면, 식 (1)을 주파수 영역에서 다음과 같이 나타낼 수 있다:

(1) can be expressed in the frequency domain as:

(2)

(2)

From the above equation (2), it can be seen that the imaginary part of the CPSD function is expressed as a series of sine functions. From this, it can be expected that the sine function by the time component (τ _n ) having a large size (A _n ) in the CPSD function of p ₁ and p ₂ , that is, highly correlated, becomes a factor causing the frequency fluctuation phenomenon.

In order to remove the trigonometric components that cause the correlation delay time between the direct sound and the reflections in the component of the acoustic intensity vector composed of a combination of CPSD functions (combination of trigonometric functions), the following two methods , CPSD linearization and CPSD filtering.

3. Compensation by CPSD Linearization

Consider the time delay τ ₀ between the signals p ₁ (t) and p ₂ (t) received through the two microphones, α is the attenuation factor by time delay, and n ₁ and n ₂ are measured by the microphone The uncorrelated noise of a signal is expressed as follows:

(3)

(3)

Let p ₁ (t) and p ₂ (t) denote the cross-correlation function and the CPSD function as follows:

(4)

(4)

(5)

(5)

(6)

(6)

(7)

(7)

Since the CPSD function between two microphones in the ideal free sound field without reflection, refraction, diffraction and scattering takes into account only the time delay (τ ₀ ), the magnitude function of Equation (6) (7) can be replaced by linearization to remove the trigonometric component due to the correlation delay time (τ _n ) between the direct sound and the reflected sound.

Here, since the response of the phase function is sensitive to the slope (? ₀ ), that is, the time lag due to the direct sound, a proper linearization interval should be selected in consideration of the interval between the microphones.

4. Compensation by CPSD filtering

(2) that the imaginary part of the CPSD function consists of a series of sine functions. Therefore, a filter capable of removing components other than the delay time (? ₀ ) due to direct sound in the time domain is applied to the frequency domain.

Since τ ₀ is close to zero, it should be a low pass filter, and the cut-off time τ _r is determined by considering the interval between the microphones.

FIR filters are used to ensure linearity of the phase. When estimating the position, IIR filter may be applied to minimize the time required. Computation time results when the filter was applied to six mutual power spectral density functions using a computer with CPU performance of Intel i7-4790K 4GHz and 16GB memory was 3.87 seconds for FIR and 0.03 seconds for IIR.

5. Compensation of deflection error according to sound source location

The error width varies depending on the angle at which the sound is incident and the frequency range. In the embodiment of the present invention, it is found out that there is a cause of the directivity of the three-dimensional acoustic intensity probe apparatus, and a method of compensating the bias error based thereon is presented.

The phase difference (ψ _αβ) between the specific direction by θ, φ any microphone α, β measuring a plane wave incident on the distance is defined as follows:

(8)

(8)

는 각각 음원과 마이크로폰 α,β 사이의 위치 벡터를 의미하고,

은 음원의 법선벡터를 의미한다.here,

Denotes a position vector between the sound source and the microphones alpha and beta, respectively,

Is the normal vector of the sound source.

,

(9,10)

,

(9, 10)

In the measurement of the 1-dimensional acoustic intensity, as in the case of suggesting a criterion for setting the frequency region of interest in consideration of the phase misalignment error in the low frequency region and the finite difference error in the high frequency region, in the measurement of the 3-dimensional acoustic intensity vector, Or the value is determined according to the phase difference depending on the distance, the frequency domain of the region of interest can be set in consideration of the phase difference _??? .

The directional coefficients, the azimuth angle, the error range and the standard deviation of the angle of incidence observed for a sound incident in a specific direction are summarized in FIG. Fig. 3 is a map showing the direction coefficient, the azimuth angle, the error range and the standard deviation of the altitude angle according to the change of kd.

If the three-dimensional acoustic intensity vector is corrected using the biased error map of FIG. 3 analyzed on the basis of the directivity of the probe with respect to the frequency (kd) value of the sound source for the same microphone interval, And a deflection error that varies depending on the incident direction of the sound wave.

6. Test results

Figure 4 shows the results of an experiment on the impact sound in an actual anechoic room condition by varying the angle of the plane at an altitude angle of 0 degrees. In FIG. 4, the green dotted line represents the actual direction of the sound source, the blue solid line represents the result before compensation, the black solid line represents the compensation result by filtering, and the red dotted line represents the compensation result by the linearization.

It can be seen that the location of the impact sound is well estimated at an azimuth error of 0.4 ° and an altitude angle error of 1.4 ° within 500-1500 Hz, which was the frequency band of interest.

The fact that the position estimation is performed well using 0.02 second single sample data suggests that the location of the sound source can be estimated in real time through the acoustic intensity and error compensation algorithm measured through four microphones.

7. Application Examples

There are many areas where sound source location technology can be applied, such as future equipment field, military equipment field, and security equipment field.

Existing commercialized products track the position of a sound source with a signal received from a myriad of microphones using a signal processing technique such as a beam forming method, and the cost of the design is increased and the physical size of the entire system is increased, , ICT equipment, and military equipment.

On the other hand, the acoustic intensity tracking apparatus according to the embodiment of the present invention significantly reduces the number of necessary microphones, reduces the physical size of the system, and improves the location tracking performance of the sound source as well as miniaturization of the system.

8. Variations

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is clearly understood that the same is by way of illustration and example only and is not to be construed as limiting the scope of the invention as defined by the appended claims. It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention.

111 to 114: a microphone
120: Cross Power Spectral Density (CPSD)
130: CPSD function compensating unit
140: Acoustic Intensity Calculator
150: Acoustic Intensity Correction

Claims (10)

Generating mutual power spectral density functions between the signals measured in the plurality of microphones;
Compensating the generated mutual power spectral density functions; And
Combining the compensated mutual power spectral density functions to produce an acoustic intensity vector,
Wherein the compensating step comprises:
And linearizing the phase functions of the generated mutual power spectral density functions.
delete The method according to claim 1,
The phase function? Is expressed by the following equation,
? = 2? t ₀ f,
Wherein τ ₀ is a time delay according to the distance between the microphones.
The method according to claim 1,
Wherein the compensating step comprises:
And the generated mutual power spectral density functions are subjected to low-pass filtering.
Generating mutual power spectral density functions between the signals measured in the plurality of microphones;
Compensating the generated mutual power spectral density functions; And
Combining the compensated mutual power spectral density functions to produce an acoustic intensity vector,
Wherein the compensating step comprises:
Pass filtering the generated mutual power spectral density functions,
The cut-off frequency time by the low-
And the distance between the microphones is determined according to the distance between the microphones.
The method according to claim 1,
The components of the acoustic intensity vector,
Wherein the power spectral density functions are calculated from imaginary parts of the mutual power spectral density functions.
The method according to claim 1,
And correcting the calculated acoustic intensity vector according to the frequency of the sound,
Wherein the correcting step comprises:
And correcting a deflection error according to an incidence direction of the sound.
The method according to claim 1,
The plurality of microphones,
And four microphones arranged at apexes of the regular tetrahedron.
The method according to claim 1,
Wherein the compensating step comprises:
Wherein the generated mutual power spectral density functions are compensated for to eliminate a bias error caused by a plurality of microphones.
A generator for generating mutual power spectral density functions between the signals measured in the plurality of microphones;
A compensation unit for compensating the generated mutual power spectral density functions; And
And a calculation unit for calculating the acoustic intensity vector by combining the compensated mutual power spectral density functions,
Wherein the compensation unit comprises:
And linearizes the phase functions of the generated mutual power spectral density functions.

Images