CN104469154A - Camera guiding device and method based on microphone array - Google Patents

Abstract

The invention relates to a camera guiding device and method based on a microphone array. The guiding device comprises a camera, the microphone array and a multichannel sound signal synchronous collection processing system, wherein the camera is provided with a control pan-tilt. The multichannel sound signal synchronous collection processing system comprises a conditioning circuit, a Codec chip, a DSP chip and a 485 interface. The conditioning circuit is used for conditioning analog signals collected by microphones. The Codec chip is used for converting the analog signals into digital signals and sending the digital signals into a DSP chip. The DSP chip is used for conducting sudden sound signal detection and sudden sound signal direction sensing and transmitting a direction sensing result to the control pan-tilt of the camera through the 485 interface according to the received digital signals, and therefore the angle of the camera is adjusted so that the camera can aim at a target to conduct shooting. The guiding device has self-adaption performance, and is wide in application range and capable of preventing the camera from rotating in an aimless mode, and meanwhile the estimation precision of the direction and the robustness of a system are high.

Description

A kind of camera guiding device based on microphone array and bootstrap technique

Technical field

The invention belongs to monitoring technique field, relate to one and utilize microphone array to detect and direction finding burst sound source, and guide camera to catch the device and method of this burst sound source.

Background technology

At present, video monitoring is an important means in monitoring field, is used widely.But camera has blind area, the visual angle of general camera is 52 °, monitoring range is limited, the normally regular rotation of rotating camera or by Artificial Control, is difficult to capture accident in time, as the riot, the warning of residential quarter vehicle theft, traffic accident etc. of public place.If these events are not at the guarded region of camera, traditional video monitoring system is just helpless.

As everyone knows, people can be positioned by the sound of ear to accident, and then adjusts visual angle in time.Use for reference the mechanism of perception of people, can load onto " ear " for camera, this " ear " is exactly microphone array.Utilize the sound of microphone array to accident to detect and direction finding, and guide camera, realize monitoring in real time the audio frequency and video of accident.

Utilize microphone array to position sound source at present, and then guide camera aligning sound source to have multiple patent.US Patent No. 6,826, is utilized 6 microphone composition solid arrays, is positioned, thus guide camera to point to sound source by time delay Direction Finding Algorithm (TDOA) to sound source in 284 B1.Chinese patent CN10156798B discloses the intelligent video director method of leading based on microphone array sound intelligence, first it carry out acquisition parameters setting to key area, example is kept in training set, then microphone array is utilized to position sound source, and the example in positioning result and training set is compared calculate, finally guide camera to take by given parameters.Chinese utility model patent CN203151672U discloses a kind of video system of sound source positioning function, and it on the hardware system oneself built, achieve auditory localization and the camera controlled on The Cloud Terrace aims at sound source.

But in these patents above, common shortcoming is: 1. do not have sound detection module, if do not add sound detection module, the aligning accident that camera can not be shot the arrow at the target, and whole system is easily disturbed by outside noise; 2. what utilize in above patent is all TDOA class sound localization method, and these class methods require that microphone interval is comparatively large, therefore cause whole device volume huge, are difficult to practicality in practice; And such Direction Finding Algorithm belongs to time-domain processing method, and the frequency range of voice signal under different scene is different, and these class methods are not selected sound frequency.

Summary of the invention

The present invention is directed to above-mentioned the deficiencies in the prior art, proposing a kind of microphone array that utilizes guides camera to catch the device and method of burst sound source, which use noise spectrum estimation technology and broadband signal Mutual coupling technology, therefore there is adaptivity, applied range, can avoid the anomic rotation of camera; Meanwhile, the estimated accuracy in direction and the robustness of system high.

To achieve these goals, the invention provides following technical scheme: a kind of camera guiding device based on microphone array, it comprises the camera with controlling The Cloud Terrace, the microphone array arranged around described camera be made up of multiple microphone and multi-channel sound signal synchronous collection treatment system, wherein, described multiple microphone is connected to described multi-channel sound signal synchronous collection treatment system by tone frequency channel wire thus is transported to by multichannel microphone signal in described multi-channel sound signal synchronous collection treatment system, described multiple channel acousto tone signal System of Synchronous Processing comprises modulate circuit, Codec chip, dsp chip and 485 interfaces, wherein, described modulate circuit is used for nursing one's health the analog signal of described multiple microphone collection, described Codec chip is used for the analog signal conditioned through described modulate circuit being converted to digital signal and sending in described dsp chip by described digital signal, described dsp chip is used for carrying out burst acoustical signal processing and the direction finding of burst acoustical signal according to received digital signal, and direction finding result is sent to the control The Cloud Terrace of described camera by described 485 interfaces, thus the visual angle adjusting described camera is taken to aim at the mark.

In addition, the present invention also provides a kind of camera bootstrap technique adopting the above-mentioned camera guiding device based on microphone array, it comprises the following steps: the first step: start described camera guiding device, by described Codec chip, the analog signal that described multiple microphone collects is converted to digital signal, and described digital signal is sent into described dsp chip; Second step: by described dsp chip according to one of them microphones to signal adaptive estimate environmental noise power; 3rd step: by described dsp chip calculate this microphones to the energy of signal and the ratio of described environmental noise power, according to this ratio in judgement whether have burst acoustical signal occur, if proceed to the 4th step; If not, proceed to second step, continue according to microphones to signal adaptive estimate environmental noise power; 4th step: by the burst sound detection result of described dsp chip according to the 3rd step, intercept out the signal segment having burst sound in described multiple microphone; 5th step: the signal of the microphone intercepted in described 4th step is carried out fast Fourier transform respectively by described dsp chip, become multiple narrow band signal, according to the feature of different application scenarioss with burst acoustical signal, choose azimuth and the angle of pitch of suitable band utilization broadband DOA estimation algorithm estimating target; 6th step: by described dsp chip, described azimuth and the angle of pitch are sent to the control The Cloud Terrace of described camera by described 485 interfaces, the visual angle that described control The Cloud Terrace adjusts described camera according to this azimuth and the angle of pitch is taken.

Further, wherein, in described second step, suppose certain microphones to signal be x (t), then according to this microphones to signal adaptive estimate environmental noise power P _nt () can be expressed as with formula

$P_n\left(t\right)=\underset{f=f_L}{\overset{f_H}{\mathrm{\Σ}}}{P}_n(t,f)=\underset{f=f_L}{\overset{f_H}{\mathrm{\Σ}}}\mathrm{\α}\left(t\right)P_n(t-1,f)+(1-\mathrm{\α}\left(t\right)){\left|X(t,f)\right|}^2$

Wherein, f _land f _hthe low-limit frequency of the signal arrived for microphones and highest frequency, P _n(t, f) is the energy of signal at frequency f place that microphones arrives, P _n(t-1, f) is the energy of signal at previous moment frequency f place that microphones arrives, and X (t, f) is the Fourier transform of the signal that microphones arrives, and α (t) is smoothing factor.

Further, wherein, in described 3rd step, this microphones to the energy of signal be the then energy of signal that arrives of this microphones and the ratio SNR=P of environmental noise power _s(t)/P _nt (), as SNR > T _htime, burst acoustical signal detected; As SNR≤T _htime, burst acoustical signal do not detected, wherein T _hfor detection threshold.

Again further, wherein, in described 5th step, described broadband DOA estimation algorithm selects frequency-domain beamforming algorithm, frequency-domain beamforming algorithm is by doing Fourier transform to received signal, time delay between different microphone is converted to the phase difference on each frequency band, and then on each frequency band, application of beam formation algorithm obtains the angle estimation on this frequency band, finally the angle estimation result on these frequency bands is averaged the angle estimation obtaining echo signal.

Again further, wherein, particularly, assuming that the signal that described microphone array receives is x _n(t), n=1,2 ..., N, wherein N is the number of microphone in microphone array,

First to the received signal do fast Fourier transform, obtain the expression X of signal on each frequency band _nf (), becomes the form of column vector by the signal indication of N number of microphone

X(f)＝[X ₁(f)，X ₂(f)，...，X _N(f)] ^T

Wherein [] ^trepresent vector transpose conversion,

Obtain covariance matrix R (f) of array signal at frequency band f place according to vectorial X (f), can be expressed as with formula

R(f)＝E[X(f)X(f) ^H]

Wherein E [] is for getting mean operation,

Utilize covariance matrix R (f) just can obtain the cost function of broad-band EDFA algorithm

Wherein θ and be respectively the angle of pitch and the azimuth of echo signal, for the steering vector of microphone array

Wherein for echo signal is to the time delay between the n-th microphone and the 1st microphone, this delay can be determined according to the shape of microphone display and position,

Finally by the angle of pitch and the azimuth that just can obtain echo signal to the search of cost function, can be expressed as with formula

Compared with prior art, tool of the present invention has the following advantages:

1, add sound detection, utilize noise spectrum estimation technology to estimate the energy of stationary noise in surrounding environment in real time, then detect burst acoustical signal according to signal to noise ratio, this detection method has adaptivity, avoids the setting of hard-threshold, applied range; Meanwhile, the anomic rotation of camera can also be avoided.

2, Direction Finding Algorithm employs broadband DOA estimation algorithm, and compared with the TDOA algorithm of time domain, the size of array can obviously reduce, and can be fixed on around camera by microphone like this, is coincide by the coordinate system of microphone array coordinate system and camera; And according to the feature of echo signal, selectively can choose the angle that some frequency band specific calculates echo signal, improve the estimated accuracy in direction and the robustness of system.

Accompanying drawing explanation

Fig. 1 is the schematic diagram of the camera guiding device based on microphone array of the present invention.

Fig. 2 is the block diagram of multi-channel sound signal synchronous collection treatment system of the present invention.

Fig. 3 is the flow chart of camera bootstrap technique of the present invention.

Fig. 4 is an exemplary application scenarios.

Embodiment

Describe the specific embodiment of the present invention in detail below in conjunction with accompanying drawing, the content of embodiment is not as the restriction to protection scope of the present invention.

The present invention relates to and utilize microphone array to guide camera to catch the device and method of burst sound source, which use noise spectrum estimation technology and broadband signal Mutual coupling technology, therefore there is adaptivity, applied range, the anomic rotation of camera can be avoided; Meanwhile, the estimated accuracy in direction and the robustness of system high.

Fig. 1 shows the schematic diagram of the camera guiding device based on microphone array of the present invention.As shown in Fig. 1, the camera guiding device based on microphone array of the present invention comprises the camera 1 with controlling The Cloud Terrace.Preferably, described camera 1 is HIVISION DS mono-2AF4262 high-speed ball-forming machine, and this camera is furnished with control The Cloud Terrace, can level 360 °, and vertical 90 ° rotate arbitrarily.This control The Cloud Terrace has control end interface, and user can by this interface adjustment camera visual angle.

In addition, the described camera guiding device based on microphone array also comprises the microphone array 2 arranged around described camera 1 and multi-channel sound signal synchronous collection treatment system 3 that are made up of multiple microphone.Wherein, described multiple microphone is connected to described multi-channel sound signal synchronous collection treatment system 3 by tone frequency channel wire thus is transported to by multichannel microphone signal in described multi-channel sound signal synchronous collection treatment system 3.

As shown in Figure 2, described multiple channel acousto tone signal System of Synchronous Processing comprises modulate circuit, Codec chip, dsp chip and 485 interfaces.Wherein, described modulate circuit is used for nursing one's health the analog signal of described multiple microphone collection, so that described Codec chip converts described analog signal to digital signal.Described Codec chip is used for the analog signal conditioned through described modulate circuit being converted to digital signal and described digital signal being sent in described dsp chip.Described dsp chip is used for carrying out burst acoustical signal processing and the direction finding of burst acoustical signal according to received digital signal, and direction finding result is sent to the control The Cloud Terrace of described camera 1 by described 485 interfaces, thus the visual angle adjusting described camera 1 is taken to aim at the mark.

As shown in Figure 3, when adopting the guiding camera of the above-mentioned camera guiding device based on microphone array, the first step starts described camera guiding device, the burst acoustical signal that may occur is obtained around by described multiple microphone, and by described Codec chip, the analog signal that described multiple microphone collects is converted to digital signal, and described digital signal is sent into described dsp chip.

Second step be by described dsp chip according to a microphones to signal adaptive estimate environmental noise power.Particularly, suppose certain microphones to signal be x (t), then according to this microphones to signal adaptive estimate environmental noise power P _nt () can be expressed as with formula

$P_n\left(t\right)=\underset{f=f_L}{\overset{f_H}{\mathrm{\Σ}}}{P}_n(t,f)=\underset{f=f_L}{\overset{f_H}{\mathrm{\Σ}}}\mathrm{\α}\left(t\right)P_n(t-1,f)+(1-\mathrm{\α}\left(t\right)){\left|X(t,f)\right|}^2$

Wherein, f _land f _hthe low-limit frequency of the signal arrived for microphones and highest frequency, P _n(t, f) is the energy of signal at frequency f place that microphones arrives, P _n(t-1, f) be that the signal that arrives of microphones is at the energy of previous moment at frequency f place, X (t, f) be the Fourier transform of the signal that microphones arrives, α (t) is smoothing factor, this coefficient according to microphones to the feature of signal suitably choose.

3rd step be by described dsp chip calculate this microphones to the energy of signal and the ratio of described environmental noise power, i.e. signal to noise ratio, according to this ratio in judgement whether have burst acoustical signal occur, if proceed to the 4th step; If not, proceed to second step, continue according to microphones to signal adaptive estimate environmental noise power.Particularly, this microphones to the energy of signal be the then energy of signal that arrives of this microphones and the ratio SNR=P of environmental noise power _s(t)/P _nt (), as SNR > T _htime, burst acoustical signal detected, enter the 4th following stepping row relax; As SNR≤T _htime, burst acoustical signal do not detected, continue to estimate environmental noise power, wherein T according to the signal adaptive of microphone _hfor detection threshold, according to this microphones to the feature of signal choose.

4th step is by the burst sound detection result of described dsp chip according to the 3rd step, intercepts out the signal segment having burst sound in described multiple microphone.

5th step is, by described dsp chip, the signal of the microphone intercepted in described 4th step is carried out fast Fourier transform respectively, become multiple narrow band signal, according to the feature of different application scenarioss with burst acoustical signal, choose azimuth and the angle of pitch of suitable band utilization broadband DOA estimation algorithm estimating target.Particularly, described broadband DOA estimation algorithm selects frequency-domain beamforming algorithm, frequency-domain beamforming algorithm is by doing Fourier transform to received signal, time delay between different microphone is converted to the phase difference on each frequency band, then on each frequency band, application of beam formation algorithm obtains the angle estimation on this frequency band, finally the angle estimation result on these frequency bands is averaged the angle estimation obtaining echo signal.Wherein, assuming that the signal that described microphone array receives is x _n(t), n=1,2 ..., N, wherein N is the number of microphone in microphone array,

First to the received signal do fast Fourier transform, obtain the expression X of signal on each frequency band _nf (), becomes the form of column vector by the signal indication of N number of microphone

X(f)＝[X ₁(f)，X ₂(f)，...，X _N(f)] ^T

Wherein [] ^trepresent vector transpose conversion,

Obtain covariance matrix R (f) of array signal at frequency band f place according to vectorial X (f), can be expressed as with formula

R(f)＝E[X(f)X(f) ^H]

Wherein E [] is for getting mean operation,

Utilize covariance matrix R (f) just can obtain the cost function of broad-band EDFA algorithm

Wherein θ and be respectively the angle of pitch and the azimuth of echo signal, for the steering vector of microphone array

Wherein for echo signal is to the time delay between the n-th microphone and the 1st microphone, this delay can be determined according to the shape of microphone display and position,

Finally by the angle of pitch and the azimuth that just can obtain echo signal to the search of cost function, can with formula

To be expressed as

6th step is, by described dsp chip, described azimuth and the angle of pitch are sent to the control The Cloud Terrace of described camera by described 485 interfaces, and the visual angle that described control The Cloud Terrace adjusts described camera according to this azimuth and the angle of pitch is taken.

After adjustment, continue to enter described second step, in order to judge whether the appearance of new burst acoustical signal, proceed monitoring and adjustment.

To illustrate that the camera bootstrap technique based on microphone array of the present invention is the adjustment at the visual angle how realizing camera with an exemplary example below.

The present embodiment carries out in the room of a 6m × 8m, and equipment is placed on room central authorities, and burst acoustical signal is sound equipment, and sound equipment plays one section of voice signal.As shown in Figure 1, camera is HIVISIONDS-2AF4262 high-speed ball-forming machine to equipment, and this camera is furnished with The Cloud Terrace, can level 360 °, and vertical 90 ° rotate arbitrarily, and this The Cloud Terrace has control end interface, and user can by this interface adjustment camera visual angle.Arrange 4 MEMS microphone around camera, constitute a round battle array, the radius of circle battle array is 74mm.4 microphones are connected in multi-channel sound signal synchronous collection treatment system by tone frequency channel wire, and the theory diagram of synchronous acquisition treatment system as shown in Figure 2.In synchronous acquisition treatment system, Codec chip is responsible for the analog signal of 4 road microphones being converted to digital signal and is sent in dsp chip, dsp chip carries out burst acoustical signal processing and the direction finding of burst acoustical signal according to the signal received, and direction finding result is sent to the The Cloud Terrace of camera by 485 interfaces, thus adjustment camera visual angle aims at the mark and takes.

As shown in Figure 4, assuming that equipment is placed on the central authorities in room, sound equipment is placed on 90 ° of equipment to the concrete layout of this example respectively, 135 °, on 180 ° of three positions, sound equipment distance equipment 2.5m, sound equipment and equipment, in same level, therefore only need in this example to consider azimuth.The concrete steps of this example are as follows:

The first step: camera guiding device is energized, microphone array gathers around acoustical signal, and the analog signal of microphone collection is converted to digital signal by the Codec chip in multi-channel sound signal synchronous collection treatment system, and sample rate is 8k.

Second step: dsp chip estimates ambient noise energy in real time according to first via signal after carrying out framing and windowing process to the 4 road signals received.

3rd step: dsp chip calculates signal to noise ratio according to the noise energy of estimation, judges whether signal to noise ratio is greater than certain threshold value.If signal to noise ratio is greater than threshold value, then proceed to the 4th step; Otherwise proceed to second step.

4th step: dsp chip, according to the signal detecting result of the 3rd step, intercepts out the data having echo signal in No. 4 microphones.

5th step: the 4 road microphone signal segmentations that dsp chip will intercept, and FFT (fast Fourier) conversion is carried out respectively to every segment data, obtain X _l(f), l=1,2 ..., L, wherein L is segmentation number, utilizes this L vector to estimate the covariance matrix of array signal method of estimation is as follows

$\hat{R}\left(f\right)=\underset{l=1}{\overset{L}{\mathrm{\Σ}}}{X}_l\left(f\right)X_l{\left(f\right)}^H$

Wherein [] ^tfor the conjugate transpose of vector.

According to the position of 4 tunnel Mikes, determine that the steering vector of array is

$a(f,\mathrm{\θ})={[e^{-j2\mathrm{\π}\mathrm{fr}\cos \left(\mathrm{\θ}\right)/c},e^{j2\mathrm{\π}\mathrm{fr}\sin \left(\mathrm{\θ}\right)/c},e^{j2\mathrm{\π}\mathrm{fr}\cos \left(\mathrm{\θ}\right)/c},e^{-j2\mathrm{\π}\mathrm{fr}\sin \left(\mathrm{\θ}\right)/c}]}^T$

Wherein r=0.074m, c=340m/s.

According to array covariance matrix the cost function J (θ) of broad-band EDFA algorithm just can be obtained with array steering vector a (f, θ) $J\left(\mathrm{\θ}\right)=\underset{f=f_L}{\overset{f_H}{\mathrm{\Σ}}}a{(f,\mathrm{\θ})}^{H}R\left(f\right)a(f,\mathrm{\θ}).$

By searching for cost function, angle corresponding to the maximum of J (θ) is exactly the direction estimation of target.

6th step: the direction estimation of target is sent to control The Cloud Terrace by 485 interfaces by dsp chip, controls cloud platform rotation adjustment camera visual angle and takes.

7th step: proceed to second step.

Sound equipment is placed in 90 ° respectively, on 135 ° and 180 ° of three positions, after throwing open sound equipment, the sensing sound equipment that camera can be correct after the delay of 1s clock.

The content of embodiment is understood for the ease of those skilled in the art and uses the present invention and describe, and does not form restriction the present invention being protected to content.Those skilled in the art, after having read content of the present invention, can carry out suitable amendment to the present invention.Protection content of the present invention is as the criterion with the content of claim.When not departing from flesh and blood and the protection range of claim, the various amendments carried out the present invention, change and replacement etc. are all within protection scope of the present invention.

Claims (6)

1. the camera guiding device based on microphone array, it comprises the camera (1) with controlling The Cloud Terrace, the microphone array (2) arranged around described camera (1) be made up of multiple microphone and multi-channel sound signal synchronous collection treatment system (3), wherein, described multiple microphone is connected to described multi-channel sound signal synchronous collection treatment system (3) by tone frequency channel wire thus is transported to by multichannel microphone signal in described multi-channel sound signal synchronous collection treatment system (3), described multiple channel acousto tone signal System of Synchronous Processing comprises modulate circuit, Codec chip, dsp chip and 485 interfaces, wherein, described modulate circuit is used for nursing one's health the analog signal of described multiple microphone collection, described Codec chip is used for the analog signal conditioned through described modulate circuit being converted to digital signal and sending in described dsp chip by described digital signal, described dsp chip is used for carrying out burst acoustical signal processing and the direction finding of burst acoustical signal according to received digital signal, and direction finding result is sent to the control The Cloud Terrace of described camera (1) by described 485 interfaces, thus the visual angle adjusting described camera (1) is taken to aim at the mark.

2. adopt a camera bootstrap technique for the camera guiding device based on microphone array according to claim 1, it comprises the following steps:

The first step: start described camera guiding device, is converted to digital signal by described Codec chip by the analog signal that described multiple microphone collects, and described digital signal is sent into described dsp chip;

Second step: by described dsp chip according to one of them microphones to signal adaptive estimate environmental noise power;

3rd step: by described dsp chip calculate this microphones to the energy of signal and the ratio of described environmental noise power, according to this ratio in judgement whether have burst acoustical signal occur, if proceed to the 4th step; If not, proceed to second step, continue according to microphones to signal adaptive estimate environmental noise power;

4th step: by the burst sound detection result of described dsp chip according to the 3rd step, intercept out the signal segment having burst sound in described multiple microphone;

5th step: the signal of the microphone intercepted in described 4th step is carried out fast Fourier transform respectively by described dsp chip, become multiple narrow band signal, according to the feature of different application scenarioss with burst acoustical signal, choose azimuth and the angle of pitch of suitable band utilization broadband DOA estimation algorithm estimating target;

6th step: by described dsp chip, described azimuth and the angle of pitch are sent to the control The Cloud Terrace of described camera by described 485 interfaces, the visual angle that described control The Cloud Terrace adjusts described camera according to this azimuth and the angle of pitch is taken.

3. camera bootstrap technique as claimed in claim 2, wherein, in described second step, suppose certain microphones to signal be x (t), then according to this microphones to signal adaptive estimate environmental noise power P _nt () can be expressed as with formula

$P_n\left(t\right)=\underset{f=f_L}{\overset{f_H}{\mathrm{\Σ}}}{P}_n(t,f)=\underset{f=f_L}{\overset{f_H}{\mathrm{\Σ}}}\mathrm{\α}\left(t\right)P_n(t-1,f)+(1-\mathrm{\α}\left(t\right)){\left|X(t,f)\right|}^2$

Wherein, f _land f _hthe low-limit frequency of the signal arrived for microphones and highest frequency, P _n(t, f) is the energy of signal at frequency f place that microphones arrives, P _n(t-1, f) is the energy of signal at previous moment frequency f place that microphones arrives, and X (t, f) is the Fourier transform of the signal that microphones arrives, and α (t) is smoothing factor.

4. camera bootstrap technique as claimed in claim 3, wherein, in described 3rd step, this microphones to the energy of signal be the then energy of signal that arrives of this microphones and the ratio SNR=P of environmental noise power _s(t)/P _nt (), as SNR > T _htime, burst acoustical signal detected; As SNR≤T _htime, burst acoustical signal do not detected, wherein T _hfor detection threshold.

5. camera bootstrap technique as claimed in claim 4, wherein, in described 5th step, described broadband DOA estimation algorithm selects frequency-domain beamforming algorithm, frequency-domain beamforming algorithm is by doing Fourier transform to received signal, time delay between different microphone is converted to the phase difference on each frequency band, then on each frequency band, application of beam formation algorithm obtains the angle estimation on this frequency band, finally the angle estimation result on these frequency bands is averaged the angle estimation obtaining echo signal.

6. camera bootstrap technique as claimed in claim 5, wherein, particularly, assuming that the signal that described microphone array receives is x _n(t), n=1,2 ..., N, wherein N is the number of microphone in microphone array,

First to the received signal do fast Fourier transform, obtain the expression X of signal on each frequency band _nf (), becomes the form of column vector by the signal indication of N number of microphone

X(f)＝[X ₁(f)，X ₂(f)，...，X _N(f)] ^T

Wherein [] ^trepresent vector transpose conversion,

Obtain covariance matrix R (f) of array signal at frequency band f place according to vectorial X (f), can be expressed as with formula

R(f)＝E[X(f)X(f) ^H]

Wherein E [] is for getting mean operation,

Utilize covariance matrix R (f) just can obtain the cost function of broad-band EDFA algorithm

Wherein θ and be respectively the angle of pitch and the azimuth of echo signal, for the steering vector of microphone array

Wherein for echo signal is to the time delay between the n-th microphone and the 1st microphone, this delay can be determined according to the shape of microphone display and position,

Finally by the angle of pitch and the azimuth that just can obtain echo signal to the search of cost function, can be expressed as with formula