CN102033223B - Method for positioning sound source by using microphone array - Google Patents

Abstract

The invention provides a method for positioning a sound source by using a microphone array. Three microphones are arranged in an equilateral triangle. The method comprises the following steps of: establishing a coordinate system, wherein an origin point of the coordinate system is coincided with the center of gravity of the equilateral triangle, and a first microphone is positioned in a longitudinal axis of the coordinate system; connecting the center of gravity of the equilateral triangle with three vertexes of the equilateral triangle, extending, and dividing an entire circumference into six equal intervals; calculating time differences between the sound source and the first microphone, between the sound source and the second microphone, and between the sound source and the third microphone; multiplying the time differences by a sound velocity to obtain a first distance difference between a distance from the sound source to the second microphone and a distance from the sound source to the first microphone and a second distance difference between a distance from the sound source to the third microphone and a distance from the sound source to the first microphone; preliminarily determining the interval in which the sound source falls relative to an azimuth angle phi of the center of gravity of the equilateral triangle according to symbols of the first distance difference and the second distance difference and comparison of absolute values of the first distance difference and the second distance difference; and approximating the distance R from the sound source to the center of gravity of the equilateral triangle and the azimuth angle phi to optimization points by a 0.618 method.

Description

Use the method for microphone array to auditory localization

Technical field

The present invention relates to the location of sound source, more particularly, relate to a kind of use microphone (MIC) array and come method auditory localization.

Background technology

Array Signal Processing has been widely used in numerous areas such as communication, radar, sonar, medical science and space flight and aviation.In recent years, along with carrying out of anti-terrorism war, various countries have developed the equipment of the position of various detection demolition points or firearms shooting point.In practical application; Because vision localization often is difficult for realizing; Thereby developed the system of some acoustics location; For example, the anti-sniper hand sound that goes back to power bar system, Israel Raphel company of the U.S. is visited side system, and the Canadian MacDonald Di Teweili company and the Canadian defence research and development council have also developed a kind of " ferret " by name miniweapon (Ferret) jointly and surveyed and positioning system.The new Vista operating system of Microsoft company exploitation is integrated to the support of MIC array, and the HDAudoi standard of Intel Company can realize 16 MIC, 32KHz sampling, and these provide good support for the acoustic signals ARRAY PROCESSING.

At present, the method for using microphone array to come sound source is positioned can be divided into two types on the whole: directed velocity of wave forms and time delay is estimated.Directed velocity of wave forms a kind of form that is based on peak power output steerable beam method; The voice signal that this method receives microphone array carries out the wave beam of filter delay and weighted sum formation particular orientation; Calculate the beam energy that each points to then, wherein prominent direction is exactly approximate to a maximum of sound source position in the search volume.In based on the time delay estimation approach, because each microphone life period of the signal that sound source is sent arrival is poor, this method is estimated this time delay according to the correlationship between each channel signal, thereby is come sound source position is estimated.

The method that directed velocity of wave forms need scan whole space in order to search for the position that can make beam power maximum, need carry out great amount of calculation, therefore is difficult to satisfy real-time requirement.

Fast based on time delay estimation approach computing velocity, can reach real-time requirement.These class methods go for as the such non-stationary broadband information source of voice, but when having only single information source, can work preferably, and locating accuracy is relevant with the estimated accuracy of time of arrival poor (TDOA).

Summary of the invention

The object of the present invention is to provide a kind of method of using microphone array to auditory localization, this method can be fast to auditory localization.

According to an aspect of the present invention; A kind of method of using microphone array to auditory localization is provided; Said microphone array is classified three microphones that constitute equilateral triangle as; Said method comprises: set up coordinate system, the initial point of said coordinate system overlaps with the center of gravity of said equilateral triangle, and first microphone in three microphones is positioned on the longitudinal axis of coordinate system; The center of gravity of equilateral triangle is connected with three summits of equilateral triangle and prolongs, thereby will be 6 the equal intervals that are divided in wholecircle week in the center of circle with the center of gravity of said equilateral triangle; Calculate sound source and arrive mistiming of first microphone in three microphones, second microphone, the 3rd microphone respectively; The mistiming of calculating multiply by velocity of sound, and it is poor to the second distance between the distance of first microphone to the distance and the sound source of first range difference between the distance of first microphone and sound source to the three microphones to the distance and the sound source of second microphone to obtain sound source; According to the order of magnitude comparison of the symbol of first range difference and second distance difference and first range difference, second distance difference, tentatively confirm the interval that sound source is fallen into respect to the position angle Φ of equilateral triangle center of gravity; Adopt 0.618 method that the distance R and the position angle Φ of sound source and equilateral triangle center of gravity approached optimized point, obtain final distance R and the position angle Φ that estimates.

According to a further aspect in the invention; A kind of method of using microphone array to auditory localization is provided; Said microphone array is classified three microphones that constitute equilateral triangle as; Said method comprises: set up coordinate system, the initial point of said coordinate system overlaps with the center of gravity of said equilateral triangle, and first microphone in three microphones is positioned on the longitudinal axis of coordinate system; Calculate sound source and arrive mistiming of first microphone in three microphones, second microphone, the 3rd microphone respectively; The mistiming of calculating multiply by velocity of sound, obtain the range difference that sound source arrives first microphone, second microphone, the 3rd microphone respectively; Set up about the Simultaneous Equations of sound source according to the triangle edges angular dependence to the range difference of each microphone; Any two equations in the Simultaneous Equations carry out simultaneous solution, obtain distance R and the sound source of sound source and the equilateral triangle center of gravity position angle Φ with respect to the equilateral triangle center of gravity.

Description of drawings

In conjunction with the drawings, from the description of following embodiment, the present invention these and/or others and advantage will become clear, and are easier to understand, wherein:

Fig. 1 is the synoptic diagram of employing equilateral triangle microphone array according to the present invention to auditory localization;

Fig. 2 is according to the employing equilateral triangle microphone array of the first embodiment of the invention process flow diagram to the method for auditory localization;

Fig. 3 is that employing time delay characteristic according to the present invention carries out the synoptic diagram that the orientation is cut apart;

Fig. 4 is according to the employing equilateral triangle microphone array of the second embodiment of the invention process flow diagram to the method for auditory localization;

Embodiment

The present invention is directed to based on the method for time delay localization of sound source and improve and innovate.In the present invention; Adopted accurate L1 correlation technique to improve speed and anti-wild value interference performance that time delay is estimated; Further adopting time delay characteristic to carry out the orientation cuts apart to improve computing velocity and to remove Local Extremum; Can adopt simultaneously 0.618 method to accelerate velocity of approch, thereby can use conventional microcomputer to solve the auditory localization problem, satisfy the speed and the accuracy requirement of practicality at tens of Milliseconds.

Below, specify principle of the present invention and embodiment with reference to accompanying drawing.In the present invention, adopt the microphone array that is made up of three microphones to come auditory localization, three microphones are arranged to equilateral triangle, that is, three microphones lay respectively on three summits of equilateral triangle.

Fig. 1 is the synoptic diagram of employing equilateral triangle microphone array according to the present invention to auditory localization, and Fig. 2 is according to the employing equilateral triangle microphone array of the first embodiment of the invention process flow diagram to the method for auditory localization.

See figures.1.and.2; In step 201, set up coordinate system, the initial point O (0 of coordinate system; The center of gravity of the equilateral triangle that 0) constitutes with the first microphone a, the second microphone b, the 3rd microphone c overlaps, and (for example first microphone a) is positioned on the longitudinal axis of coordinate system a microphone in three microphones.

In step 202, calculate the mistiming that sound source p arrives the first microphone a, the second microphone b, the 3rd microphone c respectively.

In step 203, the mistiming of calculating multiply by the speed (velocity of sound) that sound is propagated in air, obtain the range difference that sound source p arrives the first microphone a, the second microphone b, the 3rd microphone c respectively.

In the step of calculating the said mistiming, can adopt accurate L1 algorithm to estimate the said mistiming.Common related function can be thought estimating under 2 norms, and L1 is relevant to be estimating under 1 norm, the L1 relevant various character that have much at one with L2 of being correlated with.But, the L1 relevance ratio L2 easy for calculation of being correlated with, and antijamming capability is stronger.

Two voice signal x (n) at two microphone places and the L1 cross correlation function of y (n) define as follows:

$\mathrm{Rxy}\left(m\right)=\underset{n=1}{\overset{N}{\mathrm{\Σ}}}((\mathrm{abs}\left(x\left(n\right)\right)+\mathrm{abs}\left(y\left(n\right)\right))/2)-\underset{n=1}{\overset{N}{\mathrm{\Σ}}}\left(\mathrm{abs}(x\left(n\right)-y(n+m))\right)$

Wherein, n representes the sampling time point of voice signal, and N representes total sampling number of voice signal, m express time side-play amount.M is the integral multiple in SI.

Because first of following formula is constant to stationary stochastic process, therefore select maximum cross correlation value Rxy (m), be equivalent to and select minimum second, Here it is accurate L1 cross correlation function.

For the present invention; Voice signal is sampled, and two voice signal x (n) at known two microphone places change the m value to above-mentioned cross correlation function with the waveform of y (n) is basic identical but time delay is different; Make cross correlation value Rxy (m) maximum of two voice signal x (n) and y (n); That is, adjustment x (n) aligns with y (n) dislocation, and the m value of this moment promptly is the mistiming of two voice signal x (n) and y (n).In addition, it should be appreciated by those skilled in the art that in above-mentioned cross correlation function that abs () representes ABS function, the amplitude of x (n) and y (n) expression voice signal.

The method (that is, accurate L1 algorithm) of simple crosscorrelation comparison seeking time difference, just carry out in whole sample point so compare, and the optimal delay amount is not generally in whole sample point because of digitizing though can use the deferred message of all sampled points.Can use the method that improves sampling rate to improve estimated accuracy to this problem.Sampling rate improves, and causes the SI closeer, more accurately the estimated delay mistiming.But sampling rate receives hardware constraints, and memory device also receives hardware constraints.Therefore, under the situation that does not improve sampling rate, can also adopt the method for 3 quafric curve interpolation to estimate the value between two sampled points in the present invention, carry out cross-correlation method again and estimate, thereby improved estimated accuracy.

As shown in Figure 1, the center of gravity of equilateral triangle is the initial point O (0,0) of coordinate system.The length of side of equilateral triangle is L, and sound source p to the range difference of each microphone a, b, c is:

D1＝R2-R1(1)

D2＝R3-R1(2)

D3＝R3-R2(3)

The polar coordinates of sound source p be (R, Φ), R is the distance of the center of gravity of sound source p and equilateral triangle, Φ is the position angle of sound source p with respect to the center of gravity of equilateral triangle, R and Φ are unknown numbers to be asked.

The summit polar coordinates of three microphone a, b, c are respectively microphone a: (r, pi/2), microphone b: (r, 7 π/6), microphone c: (r ,-π/6).

In addition; With reference to Fig. 2, the center of gravity of each microphone a, b, c and equilateral triangle apart from r following relation is arranged:

In step 204, set up the Simultaneous Equations to the range difference of each microphone a, b, c about sound source p according to the triangle edges angular dependence.

With reference to Fig. 2, can draw according to the triangle edges angular dependence:

$R1=\sqrt{{\mathrm{<figure-callout\; id="2"\; label="R"\; filenames="HSA00000401858700031.png"\; state="\{\{state\}\}">R</figure-callout>}}^2+r^2-2\&times;R\&times;r\&times;\cos \left(\mathrm{\&Phi;}\right)}---\left(5\right)$

$\mathrm{<figure-callout\; id="2"\; label="R"\; filenames="HSA00000401858700031.png"\; state="\{\{state\}\}">R</figure-callout>}2=\sqrt{{\mathrm{<figure-callout\; id="2"\; label="R"\; filenames="HSA00000401858700031.png"\; state="\{\{state\}\}">R</figure-callout>}}^2+r^2-2\&times;R\&times;r\&times;\cos (\frac{2\mathrm{\&pi;}}{3}-\mathrm{\&Phi;})}---\left(6\right)$

$R3=\sqrt{{\mathrm{<figure-callout\; id="2"\; label="R"\; filenames="HSA00000401858700031.png"\; state="\{\{state\}\}">R</figure-callout>}}^2+r^2-2\&times;R\&times;r\&times;\cos (\mathrm{\&Phi;}+\frac{2\mathrm{\&pi;}}{3})}---\left(7\right)$

With equation (4)-(7) substitution equation (1)-(3), it is following to the Simultaneous Equations of the range difference of each microphone a, b, c to obtain sound source p:

$D1=\sqrt{{\mathrm{<figure-callout\; id="2"\; label="R"\; filenames="HSA00000401858700031.png"\; state="\{\{state\}\}">R</figure-callout>}}^2+r^2-2\&times;R\&times;r\&times;\cos (\frac{2\mathrm{\&pi;}}{3}-\mathrm{\&Phi;})}-\sqrt{{\mathrm{<figure-callout\; id="2"\; label="R"\; filenames="HSA00000401858700031.png"\; state="\{\{state\}\}">R</figure-callout>}}^2+r^2-2\&times;R\&times;r\&times;\cos \left(\mathrm{\&Phi;}\right)}---\left(8\right)$

$\mathrm{<figure-callout\; id="2"\; label="D"\; filenames="HSA00000401858700031.png"\; state="\{\{state\}\}">D</figure-callout>}2=\sqrt{{\mathrm{<figure-callout\; id="2"\; label="R"\; filenames="HSA00000401858700031.png"\; state="\{\{state\}\}">R</figure-callout>}}^2+r^2-2\&times;R\&times;r\&times;\cos (\mathrm{\&Phi;}+\frac{2\mathrm{\&pi;}}{3})}-\sqrt{{\mathrm{<figure-callout\; id="2"\; label="R"\; filenames="HSA00000401858700031.png"\; state="\{\{state\}\}">R</figure-callout>}}^2+r^2-2\&times;R\&times;r\&times;\cos \left(\mathrm{\&Phi;}\right)}---\left(9\right)$

$D3=\sqrt{{\mathrm{<figure-callout\; id="2"\; label="R"\; filenames="HSA00000401858700031.png"\; state="\{\{state\}\}">R</figure-callout>}}^2+r^2-2\&times;R\&times;r\&times;\cos (\mathrm{\&Phi;}+\frac{2\mathrm{\&pi;}}{3})}-\sqrt{{\mathrm{<figure-callout\; id="2"\; label="R"\; filenames="HSA00000401858700031.png"\; state="\{\{state\}\}">R</figure-callout>}}^2+r^2-2\&times;R\&times;r\&times;\cos (\frac{2\mathrm{\&pi;}}{3}-\mathrm{\&Phi;})}---\left(10\right)$

In step 205, any two equations in the Simultaneous Equations are carried out simultaneous solution, obtain the polar coordinates of sound source p, thereby accomplish location sound source S.

For example, any two equations in three equations (8)-(10) are carried out simultaneous solution, all can solve the value of position angle Φ and distance R, obtain sound source p polar coordinates (R, Φ).

Because substantial measurement errors can also further adopt equation (8) (9), (9) (10), (8) (10) simultaneous solution respectively, obtains three groups of position angle Φ and distance R respectively and separates, and gets the estimated value of three groups of mean values of separating as position angle Φ and distance R again.

Fig. 3 is that employing time delay characteristic according to the present invention carries out the synoptic diagram that the orientation is cut apart, and Fig. 4 is according to the employing equilateral triangle microphone array of the second embodiment of the invention process flow diagram to the method for auditory localization.

In step 401; Set up coordinate system; The center of gravity of the equilateral triangle that the initial point O (0,0) of coordinate system and the first microphone a, the second microphone b, the 3rd microphone c constitute overlaps, and (for example first microphone a) is positioned on the longitudinal axis of coordinate system a microphone in three microphones.

In order to prevent that approximate algorithm from getting into local minimum; Consider symbol and the size of mistiming,, the center of gravity of equilateral triangle is connected with three summits of equilateral triangle and prolongs in step 402; Thereby will be 6 the equal interval 1-6 that are divided in wholecircle week in the center of circle with the center of gravity of equilateral triangle, as shown in Figure 3.

In step 403, arbitrary sound source p arrives the mistiming of the first microphone a, the second microphone b, the 3rd microphone c respectively on the calculating circumference.

The accurate L1 algorithm of in first embodiment, describing and the method for 3 quafric curve interpolation can be applicable to no longer be repeated in this description here among second embodiment equally.

In step 404; The mistiming of calculating multiply by the speed (velocity of sound) that sound is propagated in air, obtain distance and the sound source p of sound source p to the second distance difference D2 between the distance of the first microphone a to the distance of the second microphone b and sound source p to first range difference D1 between the distance of the first microphone a and sound source p to the three microphone c.

In step 405,, tentatively confirm the interval (that is, position angle Φ falls into which interval of interval 1-6) that position angle Φ is fallen into according to the order of magnitude comparison of the symbol of the first range difference D1 and second distance difference D2 and D1, D2.

With reference to Fig. 3 and Fig. 4, suppose that D1 is that arbitrary sound source p deducts the range difference apart from pa gained of sound source p to microphone a, i.e. D1=pb-pa to microphone b apart from pb on the circumference; Suppose D2 be on the circumference sound source p to microphone c deduct the range difference apart from pa gained of sound source p apart from pc to microphone a, i.e. D2=pc-pa then has relation as shown in the table:

Interval	Φ	D 1	D 2	Character
					1	0°-60°	+	+	|D 1|＜|D 2|
2	60°-120°	-	+	|D 1|＜|D 2|
					3	120°-180°	-	-	|D 1|＜|D 2|
4	180°-240°	-	-	|D 1|＞|D 2|
					5	240°-300°	+	-	|D 1|＞|D 2|
6	300°-360°	+	+	|D 1|＞|D 2|

Therefore,, can tentatively confirm the interval that the position angle fell into, have single optimized point in each span, avoid local minimum thereby make according to the order of magnitude comparison of the symbol of D1, D2 and D1, D2.Reduce the codomain scope of estimating simultaneously, improve estimating speed.

In step 406, with 0.618 method bivariate (that is, distance R and position angle Φ) is approached optimized point, obtain final estimated result.

0.618 method is also referred to as Fibonacci method, 0.618 is golden section point.Can select excellent to 0.618 method of carrying out of the dull unique smallest point of two dimension.

In the present embodiment, get the initial value of distance R earlier.The initial value of distance R is chosen in the scope of 50-400 rice, is preferably 200 meters.If distance R is less than 50 meters, then sound source can be confirmed by eyesight easily.If distance R is greater than 400 meters, the data difference that then receives is very littler than, and the range information that comprises is less, is not easy to identification positioning.Because 0.618 method of carrying out to the initial value of distance R is approached, therefore, the initial value of distance R can be thought the measurement range to sound source p location.

Then; The initial value of fixed range R in the scope of the preliminary position angle Φ that confirms, adopts 0.618 method to dwindle the interval at optimized point place; And the minimum position angle Φ of approximate error, till the error of the position angle Φ that approaches for twice changes less than predetermined threshold (for example relative error is 0.01).Then, fixing this position angle Φ that approaches adopts the scope of 0.618 method decreased distance R, till the error of the distance R of approaching for twice changes less than another predetermined threshold, and the distance R that obtains approaching.Therefore, position angle Φ and distance R are approached, change up to the two error and reach in the specified threshold value, obtain results estimated.

Specifically, for position angle Φ, as stated,, can confirm tentatively position angle Φ falls into which interval of interval 1-6 according to the order of magnitude comparison of the symbol of D1, D2 and D1, D2.

Then, can be according to the minimum position angle Φ of following mode approximate error:

(a1) get position angle Φ and multiply by 0.618 for the interval upper limit S1 that confirms;

(a2), utilize the initial value of said position angle Φ and distance R to obtain distance R 1, R2, the R3 of sound source p to each microphone according to leg-of-mutton corner relation;

(a3) with the distance R that calculates 1, R2, R3 respectively divided by velocity of sound, can arrive the time of each microphone in the hope of sound source p;

(a4) calculate sound source p and arrive the mistiming t ' that the second microphone b and sound source p arrive the first microphone a _Ba1, and calculate sound source p and arrive the mistiming t ' that the 3rd microphone c and sound source p arrive the first microphone a _Ca1

(a5) utilize the sound source p that had before obtained to arrive the mistiming t that the second microphone b and sound source p arrive the first microphone a through accurate L1 cross correlation algorithm (cross correlation value) _BaAnd sound source p arrives the mistiming t that the 3rd microphone c and sound source p arrive the first microphone a _Ca, error of calculation error1=|t _Ba-t ' _Ba1|+| t _Ca-t ' _Ca1|;

(a6) get position angle Φ and multiply by (1-0.618), according to step (a2)-(a5) error of calculation error2=|t for upper limit S1 _Ba-t ' _Ba2|+| t _Ca-t ' _Ca2|; Here, t ' _Ba2And t ' _Ca2Be respectively that sound source p that position angle Φ calculates when multiply by (1-0.618) for upper limit S1 arrives sound source p that the second microphone b and sound source p arrive mistiming and the calculating of the first microphone a and arrives the mistiming that the 3rd microphone c and sound source p arrive the first microphone a;

(a7) if | error1-error2| is greater than predetermined threshold (for example can be 0.01), and error1＞error2, confirms that then lower limit S2 and the upper limit S1 of scope in the interval of confirming of position angle Φ multiply by between 0.618; If | error1-error2| is greater than said predetermined threshold, and error1＜error2, confirms that then the scope of position angle Φ multiply by between (1-0.618) and the upper limit S1 at upper limit S1;

(a8) above-mentioned steps (a1)-(a7) is repeated in the interval that is fallen into according to the position angle Φ that confirms once more, up to | error1-error2| is less than said predetermined threshold.Owing to differ very little between last two position angle Φ that get, therefore the mean value of last two position angle Φ that get one or both of is the final position angle Φ that estimates.

For example, suppose tentatively to confirm that the position angle falls into interval 1 (0 to 60 °), then getting position angle Φ is 60 ° * 0.618, is 37.08 °.Because in accurate L1 related algorithm, obtained the mistiming t that sound source p arrives each microphone _Ba, t _Ca, therefore can with initial value substitution equation (5)-(7) of position angle Φ (37.08 °) and distance R, can obtain distance R 1, R2, the R3 of sound source p according to leg-of-mutton corner relation to each microphone.The distance R that calculates 1, R2, R3 respectively divided by velocity of sound, can be arrived the time of each microphone in the hope of sound source p, and then can arrive the mistiming t ' that the second microphone b and sound source p arrive the first microphone a in the hope of sound source p _Ba1, and try to achieve sound source p and arrive the mistiming t ' that the 3rd microphone c and sound source p arrive the first microphone a _Ca1Then, error of calculation error1=|t _Ba-t ' _Ba1|+| t _Ca-t ' _Ca1|.

Then, get position angle Φ and be 60 ° * (1-0.618), be 22.92 °.Obtain error e rror2=|t according to above same mode _Ba-t ' _Ba2|+| t _Ca-t ' _Ca2|.

As | error1-error2| during, continue to search for according to 0.618 method greater than predetermined threshold (for example 0.01).When error1＞error2, the scope of the position angle Φ that asks is between 0～37.08 °, and getting position angle Φ is 37.08 ° * 0.618, calculates error1, get position angle Φ afterwards and be 37.08 ° * (1-0.618), calculate error2; Otherwise when error1＜error2, the scope of the position angle Φ that asks is between 22.92 °～60 °; Getting position angle Φ is 22.92 °+(60 °-22.92 °) * 0.618; Calculate error1, get position angle Φ afterwards and be 22.92 °+(60 °-22.92 °) * (1-0.618), calculate error2.Continue to find the solution according to above-mentioned identical mode.Till | error1-error2| was less than said predetermined threshold, the mean value of last two position angle Φ that get one or both of was the final position angle Φ that estimates.

After finally having estimated position angle Φ, the R that can adjust the distance approaches estimation.

Specifically, the initial value of distance R and null value have been confirmed an interval (0, initial value), can be according to the minimum distance R of following mode approximate error:

(b1) get distance R and multiply by 0.618 for the upper limit S3 in the interval confirmed;

(b2), utilize final position angle Φ that estimates and distance R to obtain distance R 1, R2, the R3 of sound source p to each microphone according to leg-of-mutton corner relation;

(b3) with the distance R that calculates 1, R2, R3 respectively divided by velocity of sound, can arrive the time of each microphone in the hope of sound source p;

(b4) calculate sound source p and arrive the mistiming t ' that the second microphone b and sound source p arrive the first microphone a _Ba3, and calculate sound source p and arrive the mistiming t ' that the 3rd microphone c and sound source p arrive the first microphone a _Ca3

(b5) utilize the sound source p that had before obtained to arrive the mistiming t that the second microphone b and sound source p arrive the first microphone a through accurate L1 cross correlation algorithm (cross correlation value) _BaAnd sound source p arrives the mistiming t that the 3rd microphone c and sound source p arrive the first microphone a _Ca, error of calculation error3=|t _Ba-t ' _Ba3|+| t _Ca-t ' _Ca3|;

(b6) get distance R and multiply by (1-0.618), according to step (b2)-(b5) error of calculation error4=|t for upper limit S3 _Ba-t ' _Ba4|+| t _Ca-t ' _Ca4|; Here, t ' _Ba4And t ' _Ca4Be respectively that sound source p that distance R is calculated when multiply by (1-0.618) for upper limit S3 arrives sound source p that the second microphone b and sound source p arrive mistiming and the calculating of the first microphone a and arrives the mistiming that the 3rd microphone c and sound source p arrive the first microphone a;

(b7) if | error3-error4| is greater than another predetermined threshold (for example can be 0.01), and error3＞error4, confirms that then lower limit S4 and the upper limit S3 of scope in the interval of confirming of distance R multiply by between 0.618; If | error3-error4| is greater than said another predetermined threshold, and error3＜error4, confirms that then the scope of distance R multiply by between (1-0.618) and the upper limit S3 at upper limit S3;

(b8) according to the interval of confirming once more that distance R fell into, repeat above-mentioned steps (b1)-(b7), up to | error3-error4| is less than said another predetermined threshold.Owing to differ very little between last two distance R of getting, therefore the mean value of last two distance R of getting one or both of is the final distance R of estimating.

For example, if the initial value of distance R is 200 meters, then getting distance R is 200 * 0.618, is 123.6 meters.Because in accurate L1 related algorithm, obtained the mistiming t that sound source p arrives each microphone _Ba, t _Ca, therefore can be according to leg-of-mutton corner relation, distance R 1, R2, the R3 of sound source p to each microphone can be obtained in position angle Φ substitution equation (5)-(7) of the final estimation that distance R (123.6 meters) and front have been tried to achieve.The distance R that calculates 1, R2, R3 respectively divided by velocity of sound, can be arrived the time of each microphone in the hope of sound source p, and then can arrive the mistiming t ' that the second microphone b and sound source p arrive the first microphone a in the hope of sound source p _Ba3, and try to achieve sound source p and arrive the mistiming t ' that the 3rd microphone c and sound source p arrive the first microphone a _Ca3Then, error of calculation error3=|t _Ba-t ' _Ba3|+| t _Ca-t ' _Ca3|.

Then, getting distance R is 200 * (1-0.618), is 76.4 meters.Obtain error e rror4=|t according to above same mode _Ba-t ' _Ba4|+| t _Ca-t ' _Ca4|.

As | error3-error4| during, continue to search for according to 0.618 method greater than predetermined threshold (for example 0.01).When error3＞error4, the scope of the distance R of asking is between 0～123.6 meter, and getting distance R is 123.6 * 0.618, calculates error3, and getting distance R afterwards is 123.6 * (1-0.618), calculates error4; Otherwise when error3＜error4, the scope of the distance R of asking is between 76.4～200 meters, and getting distance R is 76.4+ (200-76.4) * 0.618, calculates error3, and getting distance R afterwards is 76.4+ (200-76.4) * (1-0.618), calculates error4.Continue to find the solution according to above-mentioned identical mode.Till | error3-error4| was less than said predetermined threshold, the mean value of last two distance R of getting one or both of was the final distance R of estimating.

In addition, because in practical application, the estimation of distance R is important not as the estimation of position angle Φ; Therefore also can be after position angle Φ according to a preliminary estimate; Adopt fixing some R values (for example, 50 meters, 100 meters, 150 meters, 200 meters, 300 meters) to estimate, the minimum value of error identifying gets final product.

0.618 method can obtain approaching the result quickly than one-half method.

According to the present invention, analyzed time-delay characteristics, the method that the orientation is cut apart has been proposed, adopt 0.618 method to accelerate velocity of approch, thereby reduce the complexity of calculating, realized the algorithm of Fast estimation.To the estimation of time-delay, the relevant algorithm of accurate L1 has been proposed, reduced multiplying, improved the precision of estimating.In addition, adopted interpolation algorithm further to improve the estimated accuracy of time-delay, thereby obtained algorithm fast and accurately.

In addition, on the microphone layout, adopt the susceptibility high conformity of equilateral triangle layout to all directions.

Though the Triangular array to being made up of 3 microphones is described microphone localization method of the present invention; But also can adopt the quadrate array that constitutes by 4 microphones to realize microphone localization method of the present invention, only need the corresponding geometric relationship of adjustment to get final product.Equally, adopt the susceptibility high conformity of foursquare microphone layout to all directions.

Though the present invention is specifically described with reference to its exemplary embodiment and is shown; But will be understood by those skilled in the art that; Under the situation that does not break away from the spirit and scope of the present invention that are defined by the claims, can carry out the various changes of form and details to it.

Claims (6)

1. method of using microphone array to auditory localization, said microphone array are classified three microphones that constitute equilateral triangle as, and said method comprises:

Set up coordinate system, the initial point of said coordinate system overlaps with the center of gravity of said equilateral triangle, and first microphone in three microphones is positioned on the longitudinal axis of coordinate system;

The center of gravity of equilateral triangle is connected with three summits of equilateral triangle and prolongs, thereby will be 6 the equal intervals that are divided in wholecircle week in the center of circle with the center of gravity of said equilateral triangle;

Calculate sound source and arrive mistiming of first microphone in three microphones, second microphone, the 3rd microphone respectively;

The mistiming of calculating multiply by velocity of sound, and it is poor to the second distance between the distance of first microphone to the distance and the sound source of first range difference between the distance of first microphone and sound source to the three microphones to the distance and the sound source of second microphone to obtain sound source;

According to the order of magnitude comparison of the symbol of first range difference and second distance difference and first range difference, second distance difference, tentatively confirm the interval that sound source is fallen into respect to the position angle Φ of equilateral triangle center of gravity;

Adopt 0.618 method that the distance R and the position angle Φ of sound source and equilateral triangle center of gravity approached optimized point, obtain final distance R and the position angle Φ that estimates,

Wherein, the step of calculating the said mistiming comprises: for two the voice signal x (n) and the y (n) at any two the microphone places in said three microphones, and the definition cross correlation function $\mathrm{Rxy}\left(m\right)=\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}((\mathrm{Abs}\left(x\left(n\right)\right)+\mathrm{Abs}\left(y\left(n\right)\right))/2)-\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}\left(\mathrm{Abs}(x\left(n\right)-y(n+m))\right),$ Wherein, N representes the sampling time point of voice signal, and N representes total sampling number of voice signal, m express time side-play amount; Change the m value to said cross correlation function, make that the maximum m value of cross correlation value of said two voice signals is the mistiming of two voice signal x (n) and y (n).

2. method according to claim 1, wherein, the step that adopts 0.618 method to approach optimized point comprises:

The R that adjusts the distance gets initial value; The said initial value of fixed range R in the scope of the preliminary position angle Φ that confirms, adopts 0.618 method to dwindle the interval at optimized point place; And the minimum position angle Φ of approximate error, change less than predetermined threshold up to the error of the position angle Φ that approaches for twice;

The position angle Φ that fixedly approaches adopts the scope of 0.618 method decreased distance R, changes less than another predetermined threshold up to the error of the distance R of approaching for twice, and the distance R that obtains approaching obtains final distance R and the position angle Φ that estimates thus.

3. method according to claim 2, wherein, the step of approaching position angle Φ comprises:

(a1) get position angle Φ and multiply by 0.618 for the upper limit S1 in the interval confirmed;

(a2) according to leg-of-mutton corner relation, utilize the initial value of position angle Φ and distance R to obtain the distance of sound source to first microphone, second microphone, the 3rd microphone;

The distance that (a3) will calculate is calculated the time that sound source arrives first microphone, second microphone, the 3rd microphone respectively divided by velocity of sound;

(a4) calculate sound source and arrive the mistiming t ' that second microphone and sound source arrive first microphone _Ba1, and calculate sound source and arrive the mistiming t ' that the 3rd microphone and sound source arrive first microphone _Ca1

(a5) sound source of utilizing before the cross correlation value through voice signal to obtain arrives the mistiming t that second microphone and sound source arrive first microphone _BaAnd sound source arrives the mistiming t that the 3rd microphone and sound source arrive first microphone _Ca, error of calculation error1=|t _Ba-t ' _Ba1|+| t _Ca-t ' _Ca1|;

(a6) get position angle Φ and multiply by (1-0.618) for upper limit S1, according to step (a2) to step (a5) error of calculation error2=|t _Ba-t ' _Ba2|+| t _Ca-t ' _Ca2|, wherein, t ' _Ba2And t ' _Ca2Be respectively that sound source that position angle Φ calculates when multiply by (1-0.618) for upper limit S1 arrives sound source that second microphone and sound source arrive mistiming and the calculating of first microphone and arrives the mistiming that the 3rd microphone and sound source arrive first microphone;

(a7) if | error1-error2| is greater than predetermined threshold, and error1＞error2, confirms that then lower limit S2 and the upper limit S1 of scope in the interval of confirming of position angle Φ multiply by between 0.618; If | error1-error2| is greater than said predetermined threshold, and error1＜error2, confirms that then the scope of position angle Φ multiply by between (1-0.618) and the upper limit S1 at upper limit S1;

(a8) interval that is fallen into according to the position angle Φ that confirms once more; Repeat above-mentioned steps (a1) to step (a7); Up to | error1-error2| is less than said predetermined threshold, and the mean value of last two position angle Φ that get one or both of is the final position angle Φ that estimates.

4. method according to claim 3, wherein, the initial value of distance R and null value are confirmed an interval, the step of approaching distance R comprises:

(b1) get distance R and multiply by 0.618 for the interval upper limit S3 that confirms;

(b2), utilize final position angle Φ and the distance R of estimating to obtain sound source to first microphone, second microphone, the distance R 1 of the 3rd microphone, R2, R3 according to leg-of-mutton corner relation;

The distance that (b3) will calculate can arrive the time of first microphone, second microphone, the 3rd microphone in the hope of sound source respectively divided by velocity of sound;

(b4) calculate sound source and arrive the mistiming t ' that second microphone and sound source arrive first microphone _Ba3, and calculate sound source and arrive the mistiming t ' that the 3rd microphone and sound source arrive first microphone _Ca3

(b5) sound source of utilizing before the cross correlation value through voice signal to obtain arrives the mistiming t that second microphone and sound source arrive first microphone _BaAnd sound source arrives the mistiming t that the 3rd microphone and sound source arrive first microphone _Ca, error of calculation error3=|t _Ba-t ' _Ba3|+| t _Ca-t ' _Ca3|;

(b6) get distance R and multiply by (1-0.618) for upper limit S3, according to step (b2) to step (b5) error of calculation error4=|t _Ba-t ' _Ba4|+| t _Ca-t ' _Ca4|, wherein, t ' _Ba4And t ' _Ca4Be respectively that sound source that distance R is calculated when multiply by (1-0.618) for upper limit S3 arrives sound source that second microphone and sound source arrive mistiming and the calculating of first microphone and arrives the mistiming that the 3rd microphone and sound source arrive first microphone;

(b7) if | error3-error4| is greater than another predetermined threshold, and error3＞error4, confirms that then lower limit S4 and the upper limit S3 of scope in the interval of confirming of distance R multiply by between 0.618; If | error3-error4| is greater than said another predetermined threshold, and error3＜error4, confirms that then the scope of distance R multiply by between (1-0.618) and the upper limit S3 at upper limit S3;

(b8) according to the interval of confirming once more that distance R fell into, repeat above-mentioned steps (b1) to step (b7), up to | error3-error4| is less than said another predetermined threshold, and the mean value of last two distance R of getting one or both of is the final distance R of estimating.

5. method of using microphone array to auditory localization, said microphone array are classified three microphones that constitute equilateral triangle as, and said method comprises:

Set up coordinate system, the initial point of said coordinate system overlaps with the center of gravity of said equilateral triangle, and first microphone in three microphones is positioned on the longitudinal axis of coordinate system;

Calculate sound source and arrive mistiming of first microphone in three microphones, second microphone, the 3rd microphone respectively;

The mistiming of calculating multiply by velocity of sound, obtain the range difference that sound source arrives first microphone, second microphone, the 3rd microphone respectively;

Set up about the Simultaneous Equations of sound source according to the triangle edges angular dependence to the range difference of each microphone;

Any two equations in the Simultaneous Equations carry out simultaneous solution, obtain distance R and the sound source of sound source and the equilateral triangle center of gravity position angle Φ with respect to the equilateral triangle center of gravity,

Wherein, the step of calculating the said mistiming comprises: for two the voice signal x (n) and the y (n) at any two the microphone places in said three microphones, and the definition cross correlation function $\mathrm{Rxy}\left(m\right)=\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}((\mathrm{Abs}\left(x\left(n\right)\right)+\mathrm{Abs}\left(y\left(n\right)\right))/2)-\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}\left(\mathrm{Abs}(x\left(n\right)-y(n+m))\right),$ Wherein, N representes the sampling time point of voice signal, and N representes total sampling number of voice signal, m express time side-play amount; Change the m value to said cross correlation function, make that the maximum m value of cross correlation value of said two voice signals is the mistiming of two voice signal x (n) and y (n).

6. method according to claim 5; Also comprise: three groups of two different equations in the Simultaneous Equations carry out simultaneous solution; Obtain three groups of position angle Φ and distance R respectively and separate, get the estimated value of three groups of mean values of separating again as position angle Φ and distance R.

Images