CN106291469A - A kind of three dimensions source of sound localization method and system - Google Patents

Abstract

The present invention proposes a kind of three dimensions source of sound localization method and system, and including arranging the mike group including multiple mike, the sound-source signal utilizing multiple mike to record respectively estimates the distance of sound source；To the multiple mike combination of two arranged, any two is combined as a pair mike, calculates, for every a pair mike, the angle that sound source is between two mikes respectively, obtains the azimuth that sound source is between two mikes；For every a pair mike, the sound source distance obtained in conjunction with estimation, determine the position of a virtual source of sound in three dimensions；Selecting an observation station, the virtual source of sound determined for each pair of mike carries out Orthogonal Decomposition, calculates horizontal angle and the elevation angle of source of sound, the three-dimensional space position of location source of sound.The present invention obtains three dimensional orientation and the size thereof of sound source in real time, provides the location technology scheme of reliable sound producing body in non-visible region for the field such as safety monitoring and robot.

Description

A kind of three dimensions source of sound localization method and system

Technical field

The invention belongs to acoustical treatment field, particularly to a kind of three dimensions source of sound localization method and system.

Background technology

In the application of safety monitoring, the generation grabbing accident to be caught at any time, and around acquisition the most promptly and accurately Useful information, make and judging accurately, just can take emergency measure fast and effectively.

Relying primarily on monitoring presentation content in tradition security protection application to judge, this needs special personnel and supervises each Control point is checked, and cannot perceive the accident occurred beyond monitoring range.Ideally, in unmanned duty Keep, professional closes the eyes for rest, monitor blind spot etc. in the case of, it is possible to be automatically obtained the early warning to accident, associated monitoring Personnel can quickly locating events generation orientation, tentatively judge the information such as main body that event occurs, then look into by monitoring image See checking, thus accomplish not omit the accident of any generation.Simultaneously in terms of emergency measure execution, to the danger existed The quickly transmission of danger situation is the most indispensable.As there is scene in accident, emergency measure first passes through the modes such as broadcast and quickly passes Pass, in order to avoid there is bigger deadweight loss before relevant treatment personnel reach the spot.By judging in time, could take in time should Anxious measure, but still do not have related art scheme to occur at present.

Summary of the invention

The present invention is directed to the deficiency of prior art and equipment, it is provided that a kind of three dimensions source of sound localization method and system.

Spatial sound signal is recorded by the present invention by multiple mikes the most in the plane, with present invention offer Localization method obtains locus and the distance of source of sound.

Technical scheme provides a kind of three dimensions source of sound localization method, comprises the following steps,

Step 1, arranges the mike group including multiple mike, and the sound-source signal utilizing multiple mike to record respectively is estimated The distance of meter sound source, obtains sound source distance D；

Step 2, to the multiple mike combination of two arranged, any two is combined as a pair mike, for every a pair Mike calculates the angle that sound source is between two mikes respectively, obtains the azimuth angle alpha that sound source is between two mikes；

Step 3, for every a pair mike, it is thus achieved that after sound source is in the azimuth angle alpha between two mikes, in conjunction with estimation Sound source distance D obtained, determines the position of a virtual source of sound in three dimensions；

Step 4, selects an observation station, and the virtual source of sound determined for each pair of mike carries out Orthogonal Decomposition, calculates sound The horizontal angle in source and elevation angle, the three-dimensional space position of location source of sound.

And, in step 2, based on any pair mike, carry out the source of sound location, space of binaural cue, obtain at sound source Azimuth angle alpha between two mikes, it is achieved mode is as follows,

Including following sub-step,

Step 2.1, is recorded to spatial sound signal by two mikes, recording signal is carried out noise reduction process, Determine sound source type；

Step 2.2, carries out time-frequency conversion to the signal after noise reduction and frequency band division processes, and obtains three frequency bands, is respectively Medium and low frequency, medium-high frequency and high frequency；

Three frequency bands are extracted binaural cue parameter by step 2.3 respectively, including intensity difference at two ears, ears time difference and Ears dependency；

To each frequency band, step 2.4, determines that parameter combines respectively, be included in medium and low frequency and select ITD and IC parameter, in medium-high frequency ILD and IC parameter is selected with high frequency；

Step 2.5, according to the parameter value of parameter selected by step 2.4, determines the corresponding sound bearing of each frequency band and size, To low frequency, medium-high frequency, high frequency band respective horizontal angle α₁、α₂、α₃, corresponding sound source size is respectively β₁、β₂、β₃；

Step 2.6, according to sound source type and the orientation of each frequency band and size estimation original sound source orientation and size, it is achieved side Formula is as follows,

If estimating that the original sound source horizontal angle obtained is α, corresponding sound source size is respectively β, then the final orientation of sound source It is calculated as follows with size,

α=(α₁N₁+α₂N₂+α₃N₃)/3

β=(β₁N₁+β₂N₂+β₃N₃)/3

Wherein, weights N₁、N₂、N₃Calculation is as follows,

If n₁、n₂、n₃Represent the sub-band number that sound source covers at three frequency bands,

$N_i=\frac{n_i}{n_1+n_2+n_3}$

Wherein, i=1,2,3.

And, frequency band divides processing mode and is, 1.5kHz is with lower part for medium and low frequency district, and 1.5～6kHz parts are middle height Frequently, 6kHz above section is high frequency.

And, in step 2.5, table look-up by referring to default look-up table, or according to default functional relationship meter Calculate, determine the corresponding sound bearing of each frequency band and size.

And, mike group includes three mikes that equilateral triangle is distributed, and in step 5, observation station selects application scenarios The central point of Instrumental, overlaps with the central point of mike group,

If virtual source of sound A, B, C and true horizontal angle corresponding to source of sound V are respectively α_A、α_B、α_C、α_V, corresponding elevation angle It is respectively β_A、β_B、β_C、β_V, gain coefficient g that three virtual sources of sound are corresponding₁=g₂=g₃=g, g are virtual source of sound or actual source of sound Distance to zero；

Assuming that spherical sound field radius is unit 1, the distance of each virtual source of sound to coordinate axes initial point is 1, each point coordinates As follows:

A(x₁ y₁ z₁)=(g₁cosβ_Acosα_A g₁cosβ_Asinα_A g₁sinβ_A)

B(x₂ y₂ z₂)=(g₂cosβ_Bcosα_B g₂cosβ_Bsinα_B g₂sinβ_B)

C(x₃ y₃ z₃)=(g₃cosβ_Ccosα_C g₃cosβ_Csinα_C g₃sinβ_C)

V (x y z)=(gcos β_Vcosα_V gcosβ_Vsinα_V gsinβ_V)

V point coordinates by A, B, C tri-point coordinates be indicated as follows,

g₁cosβ_acosα_a+g₂cosβ_bcosα_b+g₃cosβ_ccosα_c=gcos β_vcosα_v

g₁cosβ_asinα_a+g₂cosβ_bsinα_b+g₃cosβ_csinα_c=gcos β_vsinα_v

g₁sinβ_a+g₂sinβ_b+g₃sinβ_c=gsin β_v

The source of sound coordinate relative to the centre of sphere is obtained according to above two formulas.

The present invention provides a kind of three dimensions source of sound alignment system accordingly, including with lower unit,

First module, for arranging the mike group including multiple mike, utilizes the sound that multiple mike is recorded respectively Source signal estimates the distance of sound source, obtains sound source distance D；

Second unit, for the multiple mike combination of two arranged, any two is combined as a pair mike, for Every a pair mike calculates the angle that sound source is between two mikes respectively, obtains the side that sound source is between two mikes Parallactic angle α；

Unit the 3rd, for for every a pair mike, it is thus achieved that after sound source is in the azimuth angle alpha between two mikes, Sound source distance D obtained in conjunction with estimation, determines the position of a virtual source of sound in three dimensions；

Unit the 4th, for selecting an observation station, the virtual source of sound determined for each pair of mike carries out Orthogonal Decomposition, Calculate horizontal angle and the elevation angle of source of sound, the three-dimensional space position of location source of sound.

And, second unit includes with lower module,

First module, for being recorded spatial sound signal by two mikes, carries out noise reduction to recording signal Process, determine sound source type；

Second module, processes for the signal after noise reduction carries out time-frequency conversion and frequency band division, obtains three frequency bands, point Wei medium and low frequency, medium-high frequency and high frequency；

Three module, for extracting binaural cue parameter respectively to three frequency bands, including intensity difference at two ears, ears time difference And ears dependency；

4th module, for each frequency band being determined, parameter combines respectively, is included in medium and low frequency and selects ITD and IC parameter, Medium-high frequency and high frequency select ILD and IC parameter；

5th module, for according to the parameter value of parameter selected by the 4th module, determine the corresponding sound bearing of each frequency band and Size, obtains low frequency, medium-high frequency, high frequency band respective horizontal angle α₁、α₂、α₃, corresponding sound source size is respectively β₁、β₂、β₃；

6th module, is used for according to sound source type and the orientation of each frequency band and size estimation original sound source orientation and size, Implementation is as follows,

If estimating that the original sound source horizontal angle obtained is α, corresponding sound source size is respectively β, then the final orientation of sound source It is calculated as follows with size,

α=(α₁N₁+α₂N₂+α₃N₃)/3

β=(β₁N₁+β₂N₂+β₃N₃)/3

Wherein, weights N₁、N₂、N₃Calculation is as follows,

If n₁、n₂、n₃Represent the sub-band number that sound source covers at three frequency bands,

$N_i=\frac{n_i}{n_1+n_2+n_3}$

Wherein, i=1,2,3.

And, frequency band divides processing mode and is, 1.5kHz is with lower part as medium and low frequency, and 1.5～6kHz parts are medium-high frequency, 6kHz above section is high frequency.

And, in the 5th module, table look-up by referring to default look-up table, or according to default functional relationship meter Calculate, determine the corresponding sound bearing of each frequency band and size.

And, mike group includes three mikes that equilateral triangle is distributed, and in the 5th module, observation station selects applied field The central point of scape Instrumental, overlaps with the central point of mike group,

If virtual source of sound A, B, C and true horizontal angle corresponding to source of sound V are respectively α_A、α_B、α_C、α_V, corresponding elevation angle It is respectively β_A、β_B、β_C、β_V, gain coefficient g that three virtual sources of sound are corresponding₁=g₂=g₃=g, g are virtual source of sound or actual source of sound Distance to zero；

Assuming that spherical sound field radius is unit 1, the distance of each virtual source of sound to coordinate axes initial point is 1, each point coordinates As follows:

A(x₁ y₁ z₁)=(g₁cosβ_Acosα_A g₁cosβ_Asinα_A g₁sinβ_A)

B(x₂ y₂ z₂)=(g₂cosβ_Bcosα_B g₂cosβ_Bsinα_B g₂sinβ_B)

C(x₃ y₃ z₃)=(g₃cosβ_Ccosα_C g₃cosβ_Csinα_C g₃sinβ_C)

V (x y z)=(gcos β_Vcosα_V gcosβ_Vsinα_V gsinβ_V)

V point coordinates by A, B, C tri-point coordinates be indicated as follows,

g₁cosβ_a cosα_a+g₂cosβ_b cosα_b+g₃cosβ_c cosα_c=gcos β_v cosα_v

g₁cosβ_a sinα_a+g₂cosβ_b sinα_b+g₃cosβ_c sinα_c=gcos β_v sinα_v

g₁sinβ_a+g₂sinβ_b+g₃sinβ_c=gsin β_v

The source of sound coordinate relative to the centre of sphere is obtained according to above two formulas.Use the three dimensions source of sound location that the present invention provides Method and system, can obtain three-dimensional space position and the size thereof of sound source, in real time for the field such as safety monitoring and robot non- Visibility region provide reliable sound producing body location technology scheme can anticipating risk, reduce loss, there is important market It is worth.

Accompanying drawing explanation

Fig. 1 is the layout drawing of three mikes of the embodiment of the present invention.

Fig. 2 is that the embodiment of the present invention determines virtual source of sound schematic diagram by azimuth and distance.

Fig. 3 is the source of sound location schematic diagram of the embodiment of the present invention.

Detailed description of the invention

Below in conjunction with accompanying drawing, technical solution of the present invention is illustrated.

Step 1, arranges the mike group including multiple mike, and the sound-source signal utilizing multiple mike to record respectively is estimated The distance of meter sound source；

Mike group includes more than three mikes, and embodiment uses preferred version, arranges three mikes and records. Mike number is the most more accurate, but cost and complexity can improve accordingly.The method of estimation of sound source distance has a lot, as existing The evaluation method of straight inverse ratio in technology, first selection one, from other mike distances and minimum mike, separate it direct Sound and echo, separate direct sound wave and the energy ratio of reflected sound by calculating, and estimates sound source distance for different environment, such as Applied environment can be divided into indoor and outdoors, can divide different types of outdoor scene further.

The preliminary sound source range estimation precision carried out due to this step is the highest, and the spacing of general mike is less, because of For the sake of this is easy to calculate, a mike near mike group switching centre position when being embodied as, can be selected to estimate, as Sound source distance D that range estimation obtains.

Step 2, to the multiple mike combination of two arranged, i.e. any two is combined as a pair mike, for each Mike is calculated the angle that sound source is between two mikes respectively, obtains the azimuth that sound source is between two mikes α；

The present embodiment puts into equilateral triangle with three mikes, and has flat with reference plane (generally ground level) OK, triangle placement perpendicular to the ground.See Fig. 1, if three mikes are respectively a, b, c, should have mutually three to a, b}, b, C}, { c, a}, the distance between mike is designated as L, if sound source is in mike, { a, b}, { b, c}, { azimuth angle alpha between c, a} is divided It is not designated as α_ab、α_bc、α_ca。

Present invention further propose that, based on any pair mike, carry out the source of sound location, space of binaural cue, obtain sound Source is in the azimuth angle alpha between two mikes, it is achieved mode is as follows,

Step 2.1, is recorded to spatial sound signal by two mikes, recording signal is carried out noise reduction process, Determine the type of sound source；

The result recorded mike to spatial sound signal according to this, first carries out noise reduction process by recording signal, To eliminate the background noise in recording arrangement and environment.The signal recorded is carried out sound source type detection, as quiet, then without sounding Source, need not follow the steps below；Such as sound source sounding, then need to detect sound source type, such as voice, shot, automobile sound etc., then enter Enter following steps.The frequency range of various acoustical signals has had substantial amounts of statistical result, when being embodied as, and can be according to various The frequency range of acoustical signal, energy variation trend etc. realize sound source type detection.

Step 2.2, carries out time-frequency conversion to the signal after noise reduction and frequency band division processes；

Signal after noise reduction is first carried out time-frequency conversion, then the frequency spectrum of audio signal is carried out frequency band division, the present embodiment Middle frequency band being divided into three parts, 1.5kHz is with the medium and low frequency of lower part, and the medium-high frequencies of 1.5～6kHz parts, 6kHz is with top The high frequency divided.

The conventional division methods of prior art is to be divided into 24 sub-bands according to people's perception characteristic, and concrete division is as follows:

The present invention is that owing to ILD plays a major role at more than 1.5kHz, ITD plays a major role at below 1.5kHz, therefore Proposing these sub-bands are merged process, preferred version is: first frequency band merges front ten sub-bands, second merging Middle nine sub-bands, last five sub-bands are as a frequency band.

Step 2.3, extracts binaural cue parameter to each frequency band, including intensity difference at two ears, ears time difference and ears phase Guan Xing；

Each subband of recording signal is carried out binaural cue extraction.Three cue parameters computational methods are a lot, it is preferable that The present embodiment calculates in the following way:

(1) extraction of intensity difference at two ears ILD clue

ILD is defined as the energy ratio of two microphone records signal corresponding subband.Shown in computational methods such as formula (1):

$ILD=10{\log }_{10}\left(\frac{{\mathrm{\Σ}}_{n=0}^{N-1}{x}_2^2\left(n\right)}{{\mathrm{\Σ}}_{n=0}^{N-1}{x}_2^2\left(n\right)}\right)---\left(1\right)$

Wherein, x₁(n) and x₂N () represents the subband signal of two microphone records signals respectively, N is subband signal length.

(2) extraction of ears time difference ITD clue

Shown in computational methods such as formula (2):

$ITD=\arg \underset{d}{max}\left\{\frac{{\mathrm{\Σ}}_{n=0}^{N-1}{x}_1\left(n\right)x_2(n+d)}{\sqrt{{\mathrm{\Σ}}_{n=0}^{N-1}{x}_1^2\left(n\right){\mathrm{\Σ}}_{n=0}^{N-1}{x}_2^2(n+d)}}\right\}---\left(2\right)$

Wherein, d then represents delay sampling number.

(3) extraction of ears dependency IC clue

Shown in calculation such as formula (3):

$IC=\underset{d}{max}\left|\frac{{\mathrm{\Σ}}_{n=0}^{N-1}{x}_1\left(n\right)x_2(n+d)}{\sqrt{{\mathrm{\Σ}}_{n=0}^{N-1}{x}_1^2\left(n\right){\mathrm{\Σ}}_{n=0}^{N-1}{x}_2^2(n+d)}}\right|---\left(3\right)$

Step 2.4, determines that the corresponding parameter of each frequency band combines；

Owing to ILD plays a major role at more than 1.5kHz, ITD plays a major role at below 1.5kHz, is therefore selecting ginseng With reference to this rule when of number, embodiment, according to frequency band dividing mode, selects ITD and IC parameter at medium and low frequency, at middle height Frequency and high frequency select ILD and IC parameter.When using other frequency band dividing mode, also should be with reference to this rule.

Step 2.5, according to parameter value, tables look-up with reference to the look-up table preset, or according to default functional relationship meter Calculate, determine orientation and the size thereof of the corresponding sound source of each frequency band, obtain low frequency, medium-high frequency, high frequency band respective horizontal angle α₁、α₂、α₃, Corresponding sound source size is respectively β₁、β₂、β₃；

After obtaining ILD, ITD value, can calculate to the corresponding relation of horizontal angle according to ILD, ILD or to search it corresponding Angle value.ILD, ITD can obtain look-up table or functional relationship by experiment with the corresponding relation of horizontal angle, thus by looking into Table or be calculated the horizontal angle of sound source.After obtaining IC, can calculate according to the corresponding relation of IC Yu sound source size or Search the size of sound source.IC can obtain look-up table or functional relationship by experiment with the corresponding relation of sound source size, thus logical Cross the size tabling look-up or being calculated sound source.In this step according to the selection of step 2.4, utilize the result of step 2.3.

The rule of the corresponding relation of ILD, ILD and horizontal angle is: ILD value change level azimuth between [0,15] exists [0,90 °], more than 15 is all 90 °；Between [0,700ms], change level azimuth is at [0,90 °], more than 700ms the most all for ITD It it is 90 °；And they are as parameter value change and are gradually increased to greatly 90 °.IC with the corresponding relation rule of sound source size is: IC exists Change sound source size between [0,1], be from completely self-contained two sound sources to point.When being embodied as, it is proposed that pre-build Corresponding look-up table, directly invokes to improve efficiency when running this method flow process.

Step 2.6, according to sound source type and the orientation of each frequency band and magnitude estimation sound bearing and size.

Owing to the frequency range of different sources of sound is different, therefore estimation sound source is different at the proportion of different frequency bands: such as The frequency range of tap-tap is 500 1000Hz, is concentrated mainly on medium and low frequency, therefore when estimating sound bearing, and medium and low frequency frequency The calculating weights of band are close to 1, and other two frequency bands are close to 0；Doorbell acoustic frequency scope is 1000 2000Hz, mainly concentrates At higher frequency band, part is at low-frequency band, therefore when estimating sound bearing, and the calculating weights desirable 0.3 of medium and low frequency frequency band, in The calculating weights desirable 0.7 of high frequency band, another frequency band desirable 0；The frequency range of honk is 1500 3000Hz, Be concentrated mainly on medium-high frequency, therefore estimate sound bearing time, the calculating weights of medium-high frequency frequency band close to 1, other two frequencies Band is close to 0.Therefore falling according to frequency range different in the overlapping region of different frequency bands, different frequency bands calculates the desirable difference of weights Value, calculation is: because the frequency of sound source may be included in the several sub-bands in three big frequency bands, use n₁、n₂、n₃Represent The sub-band number that sound source covers at three frequency bands.Then the computing formula of weights is:

(i is 1,2,3)

The concrete value of frequency band parameters weight computing can be finely tuned by those skilled in the art when actually used voluntarily, in order to fixed Position is more accurate.

The sound source horizontal angle harmony source size calculated according to different frequency bands is multiplied by their weights, finally estimates sound source Final orientation and size.

If the horizontal angle that medium and low frequency, medium-high frequency, high frequency band calculate and last calculated sound source horizontal angle are respectively α₁、α₂、α₃, α (value is between 0 degree to 180 degree), corresponding sound source size is respectively β₁、β₂、β₃, β (value between 0 to 1).Then The final orientation of sound source and the computing formula of size be:

α=(α₁N₁+α₂N₂+α₃N₃)/3

β=(β₁N₁+β₂N₂+β₃N₃)/3

Step 3, for every a pair mike, it is thus achieved that after sound source is in the azimuth angle alpha between two mikes, integrating step 1 estimates sound source distance D obtained, it may be determined that the position of a virtual source of sound in three dimensions；

As in figure 2 it is shown, this plane is to comprise two mikes vertical with the plane residing for three mikes, Y-axis is two The perpendicular bisector of individual mike, it is thus achieved that after sound source is in the azimuth angle alpha between two mikes, the sound source distance obtained in conjunction with estimation D, it may be determined that the position of a virtual source of sound in three dimensions；

If sound source is in mike, { a, b}, { b, c}, { azimuth angle alpha between c, a} is designated as α respectively_ab、α_bc、α_ca, in conjunction with Sound source distance D that estimation obtains, respectively determines the position of a virtual source of sound in three dimensions, and three virtual sources of sound are designated as A, B, C.

Step 4, selects an observation station, and the virtual source of sound determined for each pair of mike carries out Orthogonal Decomposition, calculates sound The horizontal angle in source and elevation angle, the three-dimensional space position of location source of sound；

Preferably, the central point of observation station selection application scenarios Instrumental, such as monitoring are the center of photographic head in applying Point.Further, the central point of instrument overlaps with the central point of mike group, it is simple to computing.

As it is shown on figure 3, the three of three-dimensional cartesian coordinate system axle XYZ are vertical, with the ball of the spherical sound field of 3D two-by-two The heart is that initial point (i.e. observation station) sets up three-dimensional coordinate model (each virtual source of sound is positioned on the sphere of spherical sound field), it is known that virtual Source of sound A, B, C and true horizontal angle corresponding to source of sound V are respectively α_A、α_B、α_C、α_V(value is between 0 degree to 359 degree), corresponding Elevation angle is respectively β_A、β_B、β_C、β_V(value is spent to 90 degree between-90), owing to three mikes of mike group are positive triangle Shape is distributed, the gain coefficient g that three virtual sources of sound are corresponding₁=g₂=g₃=g, g are that virtual source of sound or actual source of sound are to zero Distance.Assuming that spherical sound field radius is unit 1, then the distance of each virtual source of sound to coordinate axes initial point is 1, each point is sat Mark as follows:

$\begin{array}{c}A\left(\begin{array}{ccc}{x}_1& y_1& z_1\end{array}\right)=\left(\begin{array}{ccc}{g}_1\cos {\mathrm{\β}}_A\cos {\mathrm{\α}}_A& g_1\cos {\mathrm{\β}}_A\sin {{\mathrm{\α}}_A}_{}& g_1\sin {\mathrm{\β}}_A\end{array}\right)\\ B\left(\begin{array}{ccc}{x}_2& y_2& z_2\end{array}\right)=\begin{array}{c}\left(\begin{array}{ccc}{g}_2\cos {\mathrm{\β}}_B\cos {\mathrm{\α}}_B& g_2\cos {\mathrm{\β}}_B\sin {\mathrm{\α}}_B& g_2\sin {\mathrm{\β}}_B\end{array}\right)\end{array}\\ C\left(\begin{array}{ccc}{x}_3& y_3& z_3\end{array}\right)=\begin{array}{}\end{array}\left(\begin{array}{ccc}{g}_3\cos {\mathrm{\β}}_C\cos {\mathrm{\α}}_C& g_3\cos {\mathrm{\β}}_C\sin {\mathrm{\α}}_C& g_3\sin {\mathrm{\β}}_C\end{array}\right)\\ V\left(\begin{array}{ccc}x& y& z\end{array}\right)=\left(\begin{array}{ccc}g\cos {\mathrm{\β}}_V\cos {\mathrm{\α}}_V& g\cos {\mathrm{\β}}_V\sin {\mathrm{\α}}_V& g\sin {\mathrm{\β}}_V\end{array}\right)\end{array}---\left(3\right)$

V point coordinates can by A, B, C tri-point coordinates be indicated, be indicated respectively, have:

g₁cosβ_acosα_a+g₂cosβ_bcosα_b+g₃cosβ_ccosα_c=gcos β_vcosα_v

g₁cosβ_asinα_a+g₂cosβ_bsinα_b+g₃cosβ_csinα_c=gcos β_vsinα_v (4)

g₁sinβ_a+g₂sinβ_b+g₃sinβ_c=gsin β_v

Relative to the coordinate of the centre of sphere, the three-dimensional space position of source of sound can be positioned in conjunction with formula (3), (4) available source of sound.

Said method, for according to estimating gained original sound source position and size, controls photographic head and rotates to security protection sound source, And according to sound source distance and size adjusting focal length.

When being embodied as, method provided by the present invention can realize automatic operational process based on software engineering, it is possible to uses mould Massing mode realizes corresponding system.

The embodiment of the present invention provides a kind of three dimensions source of sound alignment system accordingly, including with lower unit,

First module, for arranging the mike group including multiple mike, utilizes the sound that multiple mike is recorded respectively Source signal estimates the distance of sound source, obtains sound source distance D；

Second unit, for the multiple mike combination of two arranged, any two is combined as a pair mike, for Every a pair mike calculates the angle that sound source is between two mikes respectively, obtains the side that sound source is between two mikes Parallactic angle α；

Unit the 3rd, for for every a pair mike, it is thus achieved that after sound source is in the azimuth angle alpha between two mikes, Sound source distance D obtained in conjunction with estimation, determines the position of a virtual source of sound in three dimensions；

Unit the 4th, for selecting an observation station, the virtual source of sound determined for each pair of mike carries out Orthogonal Decomposition, Calculate horizontal angle and the elevation angle of source of sound, position the three-dimensional space position of source of sound,

Further, second unit includes with lower module,

First module, for being recorded spatial sound signal by two mikes, carries out noise reduction to recording signal Process, determine sound source type；

Second module, processes for the signal after noise reduction carries out time-frequency conversion and frequency band division, obtains three frequency bands, point Wei medium and low frequency, medium-high frequency and high frequency；

Three module, for extracting binaural cue parameter respectively to three frequency bands, including intensity difference at two ears, ears time difference And ears dependency；

4th module, for each frequency band being determined, parameter combines respectively, is included in medium and low frequency and selects ITD and IC parameter, Medium-high frequency and high frequency select ILD and IC parameter；

5th module, for according to the parameter value of parameter selected by the 4th module, determine the corresponding sound bearing of each frequency band and Size, obtains low frequency, medium-high frequency, high frequency band respective horizontal angle α₁、α₂、α₃, corresponding sound source size is respectively β₁、β₂、β₃；

6th module, is used for according to sound source type and the orientation of each frequency band and size estimation original sound source orientation and size, Implementation is as follows,

If estimating that the original sound source horizontal angle obtained is α, corresponding sound source size is respectively β, then the final orientation of sound source It is calculated as follows with size,

α=(α₁N₁+α₂N₂+α₃N₃)/3

β=(β₁N₁+β₂N₂+β₃N₃)/3

Wherein, weights N₁、N₂、N₃Calculation is as follows,

If n₁、n₂、n₃Represent the sub-band number that sound source covers at three frequency bands,

$N_i=\frac{n_i}{n_1+n_2+n_3}$

Wherein, i=1,2,3.

Each several part implements and can be found in method part, and it will not go into details for the present invention.

Specific embodiment described herein is only to present invention spirit explanation for example.Technology neck belonging to the present invention Described specific embodiment can be made various amendment or supplements or use similar mode to replace by the technical staff in territory Generation, but without departing from the spirit of the present invention or surmount scope defined in appended claims.

Claims (10)

1. a three dimensions source of sound localization method, it is characterised in that: comprise the following steps,

Step 1, arranges the mike group including multiple mike, utilizes the sound-source signal estimation sound that multiple mike is recorded respectively The distance in source, obtains sound source distance D；

Step 2, to the multiple mike combination of two arranged, any two is combined as a pair mike, for every a pair Mike Wind calculates the angle that sound source is between two mikes respectively, obtains the azimuth angle alpha that sound source is between two mikes；

Step 3, for every a pair mike, it is thus achieved that after sound source is in the azimuth angle alpha between two mikes, obtain in conjunction with estimation Sound source distance D, determine the position of a virtual source of sound in three dimensions；

Step 4, selects an observation station, and the virtual source of sound determined for each pair of mike carries out Orthogonal Decomposition, calculates source of sound Horizontal angle and elevation angle, the three-dimensional space position of location source of sound.

Three dimensions source of sound localization method the most according to claim 1, it is characterised in that: in step 2, based on any pair wheat Gram wind, carries out the source of sound location, space of binaural cue, obtains the azimuth angle alpha that sound source is between two mikes, it is achieved mode It is as follows,

Including following sub-step,

Step 2.1, is recorded to spatial sound signal by two mikes, recording signal is carried out noise reduction process, determines Sound source type；

Step 2.2, carries out time-frequency conversion to the signal after noise reduction and frequency band division processes, and obtains three frequency bands, low in being respectively Frequently, medium-high frequency and high frequency；

Three frequency bands are extracted binaural cue parameter by step 2.3 respectively, including intensity difference at two ears, ears time difference and ears Dependency；

To each frequency band, step 2.4, determines that parameter combines respectively, be included in medium and low frequency and select ITD and IC parameter, at medium-high frequency and height Frequency selects ILD and IC parameter；

Step 2.5, according to the parameter value of parameter selected by step 2.4, determines the corresponding sound bearing of each frequency band and size, obtains low Frequently, medium-high frequency, high frequency band respective horizontal angle α₁、α₂、α₃, corresponding sound source size is respectively β₁、β₂、β₃；

Step 2.6, according to sound source type and the orientation of each frequency band and size estimation original sound source orientation and size, it is achieved mode is such as Under,

If estimating that the original sound source horizontal angle obtained is α, corresponding sound source size is respectively β, then the final orientation of sound source is with big Little it is calculated as follows,

α=(α₁N₁+α₂N₂+α₃N₃)/3

β=(β₁N₁+β₂N₂+β₃N₃)/3

Wherein, weights N₁、N₂、N₃Calculation is as follows,

If n₁、n₂、n₃Represent the sub-band number that sound source covers at three frequency bands,

$N_i=\frac{n_i}{n_1+n_2+n_3}$

Wherein, i=1,2,3.

Three dimensions source of sound localization method the most according to claim 2, it is characterised in that: frequency band divides processing mode and is, 1.5kHz is with lower part for medium and low frequency district, and 1.5～6kHz parts are medium-high frequency, and 6kHz above section is high frequency.

Three dimensions source of sound localization method the most according to claim 2, it is characterised in that: in step 2.5, by referring to presetting Look-up table table look-up, or calculate according to default functional relationship, determine the corresponding sound bearing of each frequency band and size.

5. according to three dimensions source of sound localization method described in claim 1 or 2 or 3 or 4, it is characterised in that: mike group includes Three mikes of equilateral triangle distribution, in step 5, observation station selects the central point of application scenarios Instrumental, with mike group Central point overlap,

If virtual source of sound A, B, C and true horizontal angle corresponding to source of sound V are respectively α_A、α_B、α_C、α_V, corresponding elevation angle is respectively For β_A、β_B、β_C、β_V, gain coefficient g that three virtual sources of sound are corresponding₁=g₂=g₃=g, g are that virtual source of sound or actual source of sound are to sitting The distance of mark initial point；

Assuming that spherical sound field radius is unit 1, the distance of each virtual source of sound to coordinate axes initial point is 1, and each point coordinates is as follows:

A(x₁ y₁ z₁)=(g₁cosβ_Acosα_A g₁cosβ_Asinα_A g₁sinβ_A)

B(x₂ y₂ z₂)=(g₂cosβ_Bcosα_B g₂cosβ_Bsinα_B g₂sinβ_B)

C(x₃ y₃ z₃)=(g₃cosβ_Ccosα_C g₃cosβ_Csinα_C g₃sinβ_C)

V (x y z)=(g cos β_Vcosα_V g cosβ_Vsinα_V g sinβ_V)

V point coordinates by A, B, C tri-point coordinates be indicated as follows,

g₁cosβ_acosα_a+g₂cosβ_bcosα_b+g₃cosβ_ccosα_c=g cos β_vcosα_v

g₁cosβ_asinα_a+g₂cosβ_bsinα_b+g₃cosβ_csinα_c=g cos β_vsinα_v

g₁sinβ_a+g₂sinβ_b+g₃sinβ_c=g sin β_v

The source of sound coordinate relative to the centre of sphere is obtained according to above two formulas.

6. a three dimensions source of sound alignment system, it is characterised in that: include with lower unit,

First module, for arranging the mike group including multiple mike, utilizes the sound source letter that multiple mike is recorded respectively Number estimate sound source distance, obtain sound source distance D；

Second unit, for the multiple mike combination of two arranged, any two is combined as a pair mike, for each Mike is calculated the angle that sound source is between two mikes respectively, obtains the azimuth that sound source is between two mikes α；

Unit the 3rd, for for every a pair mike, it is thus achieved that after sound source is in the azimuth angle alpha between two mikes, in conjunction with Sound source distance D that estimation obtains, determines the position of a virtual source of sound in three dimensions；

Unit the 4th, for selecting an observation station, the virtual source of sound determined for each pair of mike carries out Orthogonal Decomposition, calculates The horizontal angle of source of sound and elevation angle, the three-dimensional space position of location source of sound.

Three dimensions source of sound alignment system the most according to claim 6, it is characterised in that: second unit includes with lower module,

First module, for being recorded spatial sound signal by two mikes, carries out noise reduction process to recording signal, Determine sound source type；

Second module, processes for the signal after noise reduction carries out time-frequency conversion and frequency band division, obtains three frequency bands, be respectively Medium and low frequency, medium-high frequency and high frequency；

Three module, for three frequency bands are extracted binaural cue parameter respectively, including intensity difference at two ears, ears time difference and Ears dependency；

4th module, for each frequency band being determined, parameter combines respectively, is included in medium and low frequency and selects ITD and IC parameter, at middle height Frequency and high frequency select ILD and IC parameter；

5th module, for the parameter value according to parameter selected by the 4th module, determines the corresponding sound bearing of each frequency band and size, Obtain low frequency, medium-high frequency, high frequency band respective horizontal angle α₁、α₂、α₃, corresponding sound source size is respectively β₁、β₂、β₃；

6th module, for according to sound source type and the orientation of each frequency band and size estimation original sound source orientation and size, it is achieved Mode is as follows,

If estimating that the original sound source horizontal angle obtained is α, corresponding sound source size is respectively β, then the final orientation of sound source is with big Little it is calculated as follows,

α=(α₁N₁+α₂N₂+α₃N₃)/3

β=(β₁N₁+β₂N₂+β₃N₃)/3

Wherein, weights N₁、N₂、N₃Calculation is as follows,

If n₁、n₂、n₃Represent the sub-band number that sound source covers at three frequency bands,

$N_i=\frac{n_i}{n_1+n_2+n_3}$

Wherein, i=1,2,3.

Three dimensions source of sound alignment system the most according to claim 7, it is characterised in that: frequency band divides processing mode and is, 1.5kHz is with lower part as medium and low frequency, and 1.5～6kHz parts are medium-high frequency, and 6kHz above section is high frequency.

Three dimensions source of sound alignment system the most according to claim 7, it is characterised in that: in the 5th module, by referring in advance If look-up table table look-up, or calculate according to default functional relationship, determine the corresponding sound bearing of each frequency band and size.

10. according to three dimensions source of sound alignment system described in claim 6 or 7 or 8 or 9, it is characterised in that: mike group includes Three mikes of equilateral triangle distribution, in the 5th module, observation station selects the central point of application scenarios Instrumental, with mike The central point of group overlaps,

If virtual source of sound A, B, C and true horizontal angle corresponding to source of sound V are respectively α_A、α_B、α_C、α_V, corresponding elevation angle is respectively For β_A、β_B、β_C、β_V, gain coefficient g that three virtual sources of sound are corresponding₁=g₂=g₃=g, g are that virtual source of sound or actual source of sound are to sitting The distance of mark initial point；

Assuming that spherical sound field radius is unit 1, the distance of each virtual source of sound to coordinate axes initial point is 1, and each point coordinates is as follows:

A(x₁ y₁ z₁)=(g₁cosβ_Acosα_A g₁cosβ_Asinα_A g₁sinβ_A)

B(x₂ y₂ z₂)=(g₂cosβ_Bcosα_B g₂cosβ_Bsinα_B g₂sinβ_B)

C(x₃ y₃ z₃)=(g₃cosβ_Ccosα_C g₃cosβ_Csinα_C g₃sinβ_C)

V (x y z)=(g cos β_Vcosα_V g cosβ_Vsinα_V g sinβ_V)

V point coordinates by A, B, C tri-point coordinates be indicated as follows,

g₁cosβ_acosα_a+g₂cosβ_bcosα_b+g₃cosβ_ccosα_c=g cos β_vcosα_v

g₁cosβ_asinα_a+g₂cosβ_bsinα_b+g₃cosβ_csinα_c=g cos β_vsinα_v

g₁sinβ_a+g₂sinβ_b+g₃sinβ_c=g sin β_v

The source of sound coordinate relative to the centre of sphere is obtained according to above two formulas.