CN113702900A - Planar array sub-band weight optimization broadband beam forming method and device - Google Patents

Planar array sub-band weight optimization broadband beam forming method and device Download PDF

Info

Publication number
CN113702900A
CN113702900A CN202110838291.1A CN202110838291A CN113702900A CN 113702900 A CN113702900 A CN 113702900A CN 202110838291 A CN202110838291 A CN 202110838291A CN 113702900 A CN113702900 A CN 113702900A
Authority
CN
China
Prior art keywords
frequency
coefficient
calculating
weight coefficient
chebyshev
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202110838291.1A
Other languages
Chinese (zh)
Other versions
CN113702900B (en
Inventor
须彬彬
杨宏
佘超
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
First Research Institute of Ministry of Public Security
Original Assignee
First Research Institute of Ministry of Public Security
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by First Research Institute of Ministry of Public Security filed Critical First Research Institute of Ministry of Public Security
Priority to CN202110838291.1A priority Critical patent/CN113702900B/en
Publication of CN113702900A publication Critical patent/CN113702900A/en
Application granted granted Critical
Publication of CN113702900B publication Critical patent/CN113702900B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S3/00—Direction-finders for determining the direction from which infrasonic, sonic, ultrasonic, or electromagnetic waves, or particle emission, not having a directional significance, are being received
    • G01S3/80—Direction-finders for determining the direction from which infrasonic, sonic, ultrasonic, or electromagnetic waves, or particle emission, not having a directional significance, are being received using ultrasonic, sonic or infrasonic waves
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S3/00—Direction-finders for determining the direction from which infrasonic, sonic, ultrasonic, or electromagnetic waves, or particle emission, not having a directional significance, are being received
    • G01S3/80—Direction-finders for determining the direction from which infrasonic, sonic, ultrasonic, or electromagnetic waves, or particle emission, not having a directional significance, are being received using ultrasonic, sonic or infrasonic waves
    • G01S3/86—Direction-finders for determining the direction from which infrasonic, sonic, ultrasonic, or electromagnetic waves, or particle emission, not having a directional significance, are being received using ultrasonic, sonic or infrasonic waves with means for eliminating undesired waves, e.g. disturbing noises
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Abstract

The invention discloses a method and a device for forming a broadband wave beam by optimizing a planar array frequency-division weight. The method comprises the following steps: s1, acquiring the spatial position of the array and the incoming wave direction information of the target sound source, acquiring an environmental sound signal in real time, and converting the environmental sound signal into a frequency domain signal; s2, calculating a variable diagonal loading coefficient and a weight coefficient formed by corresponding super-directional beams; s3, calculating a weight coefficient formed by the planar Chebyshev beam, and determining a directional factor value interval formed by the constant beam width beam; s4, determining a constant directivity factor value, dividing a frequency domain into a plurality of frequency intervals, and respectively calculating wideband beam weight coefficients optimized by frequency division bands to update the weight coefficients; s5, adopting the updated beam weight coefficient to perform beam forming calculation on the frequency domain signal obtained currently in S1, and if the incoming wave direction of the target sound source pointed by the beam is not changed, repeatedly executing S5, otherwise executing S1-S4, calculating a new weight coefficient and updating the new weight coefficient for S5.

Description

Planar array sub-band weight optimization broadband beam forming method and device
Technical Field
The invention relates to a broadband beam forming method for optimizing a planar array sub-band weight, and also relates to a corresponding broadband beam forming device, belonging to the technical field of acoustics.
Background
With the development of audio technology, microphone array beam forming technology has been widely used in the fields of sound source localization, speech enhancement, and sound field analysis. The beamforming method may be classified into a data-independent beamforming method and a statistical optimal beamforming method according to whether it is related to a received signal. The weighting coefficients of the data-independent beamforming method are deterministic, independent of the received signal, only of the array characteristic properties; the statistical optimal beam forming method is related to the received signals and the characteristics of the microphone array, and the weight coefficient is optimized in a self-adaptive processing mode.
The quasi-regularization super-directional broadband beam forming method is taken as a typical statistical optimal beam forming method, combines a super-directional beam forming method and a delay sum beam forming method, can balance white noise gain and directivity gain of a microphone array, controls directivity of broadband super-directional beam forming by using an array gain parameter, and improves related performance of microphone array processing.
However, the regularization-like super-directional broadband beam forming method still has the following disadvantages in the technical implementation and practical application processes:
(1) good array gain can be obtained in the middle frequency band, but ideal performance cannot be obtained in the low frequency band or the high frequency band, and if the array gain control parameter is not selected well, the directivity and the output signal-to-noise ratio of the microphone array are also deteriorated.
(2) Fixed diagonal loading coefficients are adopted for different frequency points, and a method for solving the diagonal loading coefficient value cannot be provided, so that if the diagonal loading coefficients are not properly selected, the beam forming performance is seriously influenced.
(3) There is a need for more uniform suppression of interference noise in different directions outside the main lobe, and the method is not based on a planar array with a large number of channels, and lacks of practical application methods.
Disclosure of Invention
The invention provides a method for optimizing broadband beam forming by using a planar array frequency-division weight value.
Another technical problem to be solved by the present invention is to provide a wideband beamforming apparatus with optimized planar array sub-band weights.
In order to achieve the purpose, the invention adopts the following technical scheme:
according to a first aspect of the embodiments of the present invention, there is provided a wideband beamforming method with optimized weights for planar array sub-bands, comprising the following steps:
step S1, obtaining the space position of the plane microphone array and the incoming wave direction information of the target sound source, collecting the environmental sound signal in real time by using the plane microphone array, and converting the environmental sound signal into a corresponding frequency domain signal;
step S2, calculating a variable diagonal loading coefficient and a weight coefficient formed by a corresponding super-directional beam according to the requirement on noise suppression;
step S3, calculating a weight coefficient formed by the planar tangential ratio Schonflow beam according to the requirement on the beam width or the side lobe suppression degree, and determining the directional factor value interval formed by the constant beam width beam;
step S4, determining a constant directivity factor value according to the requirements for white noise suppression and directivity noise suppression, dividing the frequency range into a plurality of frequency intervals, and respectively calculating wideband beam weight coefficients optimized by frequency bands to update the weight coefficients;
and S5, performing beam forming calculation on the frequency domain signal obtained currently in the step S1 by adopting the updated broadband beam weight coefficient, obtaining a time domain signal through inverse Fourier transform, repeatedly executing the step S5 if the incoming wave direction of the target sound source pointed by the beam is not changed, otherwise executing the steps S1-S4, calculating a new weight coefficient and updating the new weight coefficient for the step S5.
Preferably, the step S2 includes the following sub-steps:
step S21, according to the factSetting white noise gain floor parameter and selecting middle and low frequency interval [ f ] based on the noise suppression requirementa,fb]Determining the corresponding frequency point sequence number kaAnd kb；
Step S22, selecting two diagonal loading coefficient reference values, and respectively calculating frequency point kaAnd kbA reference weight coefficient of (c);
step S23, respectively calculating frequency point kaAnd kbInitial value of the diagonal loading coefficient;
step S24, adjusting the diagonal loading coefficient reference value, and circularly iterating the steps S22 and S23 until the frequency point k obtained by the last iteration calculationaAnd kbThe difference between the white noise gain decibel value corresponding to the diagonal loading coefficient and the target white noise gain base limit decibel value is smaller than the minimum error threshold, the iteration process is ended, and the frequency point k is obtainedaAnd kbUpper diagonal loading factor alpha (k)a) And alpha (k)b)；
Step S25, according to the diagonal loading coefficient alpha (k)a) And alpha (k)b) Calculating a variable diagonal loading coefficient alphaVSDB(k)：
Figure BDA0003177929780000031
And step S26, calculating the weight coefficient of the variable diagonal loading super-directional beam forming at all frequency points.
Preferably, in step S22 and step S26, the reference weight coefficients at each frequency point are calculated according to the following formula;
Figure BDA0003177929780000032
in the above formula, gammad(k)+α(k)IL＝Γα(k) Representing diagonally loaded pseudo-coherent matrix at frequency point k, alpha (k) representing diagonal loading coefficient, ILIs an identity matrix; gamma-shapedd(k) Representing a pseudo-coherent matrix of an L x L dimensional microphone array in a diffuse noise field, the Zeta column of the L-th row in the matrix having an element [ gamma ]d(k)]l,ζ＝sin(2πfkdl,ζ/c)/(2πfkdl,ζAnd/c) (L-0, 1, L-1, ζ -0, 1, L-1), wherein the equivalent distance in the incoming wave direction between any microphones is determined
Figure RE-GDA0003311890550000033
(ml,nl) And (m)ζ,nζ) Two-dimensional coordinates corresponding to the microphone numbers l and ζ are respectively expressed, θ represents a pitch angle from the direction of the plane wave of the target sound source to the coordinates,
Figure RE-GDA0003311890550000034
representing the horizontal angle from the direction of the plane wave of the target sound source to the coordinate;
Figure RE-GDA0003311890550000035
a steering vector formed by representing the direction of the target sound source,
Figure RE-GDA0003311890550000036
wave numbers of incoming waves of a target sound source in an x axis and a y axis respectively, c is sound propagation speed in the environment, (x is the sound propagation speed of the target sound source in the environmentl,yl) Representing two-dimensional coordinates of the target audio source.
Preferably, the step S23 includes the following sub-steps:
step S230, respectively calculating frequency point kaAnd kbWhite noise gain decibel value formed by the super-directional beam corresponding to the reference weight coefficient;
step S231, respectively calculating frequency point kaAnd kbUpper diagonal loading factor initial value.
Preferably, in step S230: calculating the frequency point k according to the following formulaaAnd kbWhite noise gain decibel value formed by the super-directional beam corresponding to the reference weight coefficient;
Figure BDA0003177929780000041
in the above formula, WSDB,dl(k) The white noise gain formed by the super-directional beam corresponding to the reference weight coefficient at the frequency point k is represented and obtained according to the following formula:
Figure BDA0003177929780000042
where H is the conjugate transpose flag.
Preferably, in step S231, the frequency point k is obtained according to the following formulaaAnd kbInitial value of the diagonal loading coefficient;
Figure BDA0003177929780000043
in the above formula, α1And alpha2Representing the reference value of the diagonal loading coefficient,
Figure BDA0003177929780000044
representing the target white noise gain floor decibel value.
Preferably, the step S3 includes the following sub-steps:
step S31, calculating the Chebyshev coefficients of the dimensions of the x and y axes respectively, and establishing the Chebyshev coefficient vector of the rectangular plane array
Step S32, calculating a weight coefficient formed by the planar Chebyshev beam;
step S33, calculating directivity factor decibel values formed by Chebyshev beams and variable diagonal loading super-directional beams at all frequency points, and respectively selecting a maximum value from the directivity factor decibel values formed by the two beams to form a directivity factor value interval;
preferably, in step S32, the frequency f of the chebyshev beamformer is calculated according to the following formulakWeight coefficient w ofCHEB(k) Comprises the following steps:
wCHEB(k)＝diag{h}v＝ΛLv(k)
in the above formula, ΛLExpressed as L-dimensional Chebyshev coefficient, andthe quantity h is a diagonal matrix of matrix diagonal elements; v (k) represents a guide vector formed by the direction of the target sound source.
Preferably, in step S33, according to the following formula, calculating the decibel values of the directivity factor formed by the chebyshev beam at all frequency points K (K is greater than or equal to 0 and less than or equal to K);
Figure BDA0003177929780000051
in the above formula, DCHEB(k) The directivity factor of Chebyshev beam forming at all frequency points K (K is more than or equal to 0 and less than or equal to K) is obtained according to the following formula:
Figure BDA0003177929780000052
in the above formula, h represents an L-dimensional Chebyshev coefficient vector, ΛLRepresenting a diagonal matrix, Γ, having as matrix diagonal elements the L-dimensional Chebyshev coefficient vector hd(k) Representing a pseudo-coherent matrix of an L multiplied by L dimensional microphone array in a diffused noise field, H is a conjugate transpose mark, and v (k) represents a steering vector formed by the direction of a target sound source;
calculating the directivity factor decibel value of variable diagonal loading super-directional beam forming at all frequency points K (K is more than or equal to 0 and less than or equal to K) according to the following formula;
Figure BDA0003177929780000053
in the above formula, DVSDB(k) Expressing the directivity factor formed by variable diagonal loading super-directional beam at all frequency points K (K is more than or equal to 0 and less than or equal to K), and obtaining the directivity factor according to the following formula;
Figure BDA0003177929780000054
in the above formula, v (k) represents a guide vector formed by the direction of the target sound source, Γα(k) At the representation frequency point kThe pseudo-coherence matrix is loaded diagonally.
Preferably, the step S4 includes the following sub-steps:
step S41, determining a directivity factor according to the requirements for white noise and directivity noise suppression, and dividing a frequency interval based on the number of intersections of the determined directivity factor and a directivity factor curve formed by variable diagonal loading super-directional beam and the corresponding frequency point positions;
and step S42, controlling the weight coefficient formed by the broadband wave beam by utilizing the array gain adjustment parameter in a sub-band manner, and updating the weight coefficient.
Preferably, the weight coefficient of the frequency point corresponding to the constant beam width frequency band interval is calculated according to the following formula;
Figure BDA0003177929780000061
in the above formula, gammaα(k) Represents a diagonally loaded pseudo-coherent matrix at a frequency point k, beta (k) represents a gain adjustment parameter of the matrix, and ΛLA diagonal matrix in which L-dimensional chebyshev coefficient vector h is a matrix diagonal element, and v (k) a steering vector formed by the directions of the target sound sources.
According to a second aspect of embodiments of the present invention, there is provided a planar array sub-band weight optimization wideband beamforming apparatus, comprising a processor and a memory, the processor reading a computer program or instructions in the memory for performing the following operations:
step S1, obtaining the space position of the plane microphone array and the incoming wave direction information of the target sound source, collecting the environmental sound signal in real time by using the plane microphone array, and converting the environmental sound signal into a corresponding frequency domain signal;
step S2, calculating a variable diagonal loading coefficient and a weight coefficient formed by a corresponding super-directional beam according to the requirement on noise suppression;
step S3, calculating a weight coefficient formed by the planar tangential ratio Schonflow beam according to the requirement on the beam width or the side lobe suppression degree, and determining the directional factor value interval formed by the constant beam width beam;
step S4, determining a constant directivity factor value according to the requirements for white noise suppression and directivity noise suppression, dividing the frequency range into a plurality of frequency intervals, and respectively calculating wideband beam weight coefficients optimized by frequency bands to update the weight coefficients;
and S5, performing beam forming calculation on the frequency domain signal obtained currently in the step S1 by adopting the updated broadband beam weight coefficient, obtaining a time domain signal through inverse Fourier transform, repeatedly executing the step S5 if the incoming wave direction of the target sound source pointed by the beam is not changed, otherwise executing the steps S1-S4, calculating a new weight coefficient and updating the new weight coefficient for the step S5.
Compared with the prior art, the method and the device for forming the wideband beam by optimizing the planar array sub-band weight have the following characteristics:
(1) by utilizing a sub-band weight coefficient optimization strategy, frequency intervals are automatically divided according to the requirements on directivity and noise reduction in practical application, and the white noise and directivity interference noise suppression capability of the microphone array and the signal output signal-to-noise ratio are effectively improved.
(2) By calculating the variable diagonal loading coefficient on the frequency, the array gain performance on different frequency points is obviously improved, the problem of low-frequency white noise amplification is solved, and the practical value is high.
(3) The advantages of Chebyshev beam forming and super-directional beam forming are fused, the beam forming side lobe level is better controlled, the constant beam wide beam directivity on a good wide frequency range is achieved, and the effectiveness and the robustness of the beam forming method are improved.
(4) For the planar microphone array with a large number of channels, the requirement on the real-time performance of the operation is obviously reduced, and the design is simple and easy to apply and realize.
Drawings
Fig. 1 is a flowchart of a method for forming wideband beams by using a planar array with a sub-band weight optimization according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a spatial topology structure of a planar rectangular microphone array in the planar array sub-band weight optimization broadband beam forming method according to the embodiment of the present invention;
FIG. 3 is a block diagram illustrating a method for forming wideband beams with optimized weights in planar arrays according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of white noise gain comparison between a planar array sub-band weight optimized wideband beamforming method and a conventional beamforming method according to an embodiment of the present invention;
fig. 5 is a schematic diagram illustrating a directional factor comparison between a plane array sub-band weight optimization wideband beamforming method and a conventional beamforming method according to an embodiment of the present invention;
fig. 6a to fig. 6c are schematic diagrams illustrating a comparison of beam responses with frequency and angle changes between the planar array sub-band weight optimization wideband beamforming method and the delay-sum beamforming method according to the embodiment of the present invention;
fig. 7a to 7d are schematic diagrams illustrating comparison of output signal waveforms of the planar array sub-band weight optimization wideband beamforming method according to the embodiment of the present invention and the conventional beamforming method;
fig. 8 is a schematic structural diagram of a planar array sub-band weight optimization wideband beam forming apparatus according to an embodiment of the present invention.
Detailed Description
The technical contents of the present invention will be further described in detail with reference to the accompanying drawings and specific embodiments.
As shown in fig. 1, an embodiment of the present invention provides a wideband beamforming method with plane array and subband weight optimization, including the following steps:
and step S1, obtaining the spatial position of the planar microphone array and the incoming wave direction information of the target sound source, collecting the environmental sound signals in real time by using the planar microphone array, and converting the environmental sound signals into corresponding frequency domain signals.
A planar rectangular microphone array with array elements uniformly distributed by using a spatial topological structure according to a preset sampling frequency fsThe method comprises the steps of collecting an environment sound signal in real time, wherein the microphone array is an L-dimensional microphone array consisting of L-NxM omnidirectional microphones (also called array elements). It is composed ofIn fig. 2, the array elements of the L-dimensional microphone array are uniformly distributed on the x-y two-dimensional plane, N is the number of the L-dimensional microphone array in the x-axis direction, M is the number of the L-dimensional microphone array in the y-axis direction, and the distances between the array elements in the x-axis and the y-axis are dxAnd dy. With the center of the L-dimensional microphone array as the coordinate system origin, an arbitrary array element may be represented by a serial number L or (M, N), where L is 0,1,., L-1, and L is M × N + N, N is 0,1,., N-1, M is 0,1,., M-1, M represents the row number of the array element, and N represents the column number of the array element. The spatial position coordinate of any array element l or (m, n) in the coordinate system is then
Figure BDA0003177929780000081
I.e. spatial position information of the microphone array.
As shown in fig. 2, the information of the incoming wave direction of the target sound source existing in the space includes the pitch angle θ and the horizontal angle from the plane wave direction of the target sound source to the coordinates
Figure BDA0003177929780000082
As shown in fig. 3, any array element L in the L-dimensional microphone array is at a predetermined sampling frequency fsThe environmental sound signals collected in real time are subjected to digital sampling and short-time Fourier transform to obtain frequency domain signals, and the frequency domain signals comprise frequency domain signals on a plurality of frequencies. Wherein the array element is at an arbitrary frequency fkFrequency domain signal X ofl(k) K is the frequency fk＝kfs/NFFTThe frequency point sequence number of (1), the collected environment sound signal is a real signal, and the effective frequency range is [0, f ]s/2]And the frequency point sequence number ranges are as follows: k0 is less than or equal to K, wherein K is NFFT/2，NFFTIs the fourier transform sequence length. Because all array elements in the L-dimensional microphone array are at the frequency fkThe frequency domain signal above is denoted as X (k) ═ X0(k),...,Xl(k),...,XL-1(k)]TTherefore, frequency domain signals of all array elements in the L-dimensional microphone array at various frequencies can be obtained.
And step S2, calculating the variable diagonal loading coefficient and the corresponding weight coefficient for forming the super-directional beam according to the requirement on noise suppression.
The method comprises the following substeps:
step S21, setting white noise gain floor parameter according to the noise suppression requirement of practical application, and selecting middle and low frequency interval [ fa,fb]Determining the corresponding frequency point sequence number kaAnd kb。
Setting a white noise gain floor parameter including a target white noise gain floor value W according to the noise suppression requirement of practical applicationlimOr target white noise gain floor decibel value
Figure BDA0003177929780000091
Selecting a middle-low frequency interval [ fa,fb]When, 50Hz < f is generally selecteda＜fbLess than 2000Hz, corresponding frequency point with the serial number kaAnd kb. The problem of white noise amplification in the middle and low frequency interval can be solved by using a variable coefficient diagonal loading method.
Step S22, selecting two diagonal loading coefficient reference values, and respectively calculating frequency point kaAnd kbReference weight coefficient of (c).
Selecting two diagonal loading coefficient reference values alpha1And alpha2As an initial diagonal loading factor reference value, generally 10 is selected-6＜α1＜α2Less than 0.05; calculating the frequency point k according to the following formulaaReference weight coefficient of (2)
Figure BDA0003177929780000093
And kbReference weight coefficient of
Figure BDA0003177929780000094
Figure BDA0003177929780000101
In the above formula, gammad(k)+α(k)IL＝Γα(k) Representing diagonally loaded pseudo-coherent matrix at frequency point k, alpha (k) representing diagonal loading coefficient, ILIs an identity matrix. Gamma-shapedd(k) Representing a pseudo-coherent matrix of an L x L dimensional microphone array in a diffuse noise field, the Zeta column of the L-th row in the matrix having an element [ gamma ]d(k)]l,ζ＝sin(2πfkdl,ζ/c)/(2πfkdl,ζAnd/c) (L-0, 1, L-1, ζ -0, 1, L-1), wherein the equivalent distance in the incoming wave direction between any microphones is determined
Figure RE-GDA0003311890550000102
(ml,nl) And (m)ζ,nζ) Two-dimensional coordinates corresponding to the microphone numbers l and ζ are respectively expressed, θ represents a pitch angle from the direction of the plane wave of the target sound source to the coordinates,
Figure RE-GDA0003311890550000103
the horizontal angle from the direction of the plane wave of the target sound source to the coordinate is shown.
Figure RE-GDA0003311890550000104
A steering vector formed by representing the direction of the target sound source,
Figure RE-GDA0003311890550000105
wave numbers of incoming waves of a target sound source in an x axis and a y axis respectively, c is sound propagation speed in the environment, (x is the sound propagation speed of the target sound source in the environmentl,yl) Representing two-dimensional coordinates of the target audio source.
Therefore, according to equation (1), k is equal to ka，α(k)＝α1(ka)＝α1To obtain the frequency point kaA reference weight coefficient of
Figure BDA0003177929780000106
K is equal to ka，α(k)＝α2(ka)＝α2To obtain the frequency point kaAnother reference weight coefficient of
Figure BDA0003177929780000107
K is equal to kb，α(k)＝α1(kb)＝α1The frequency point k can be obtainedbA reference weight coefficient of
Figure BDA0003177929780000111
K is equal to kb，α(k)＝α2(kb)＝α2The frequency point k can be obtainedbAnother reference weight coefficient of
Figure BDA0003177929780000112
Step S23, respectively calculating frequency point kaAnd kbUpper diagonal loading factor initial value.
The method comprises the following substeps:
step S230, respectively calculating frequency point kaAnd kbAnd the white noise gain decibel value formed by the super-directional beam corresponding to the reference weight coefficient.
The method comprises the steps that white noise gain formed by a diagonally-loaded super-directional beam is used for evaluating the white noise suppression capability, the physical quantity represents the array gain of a microphone array beam former under the condition of spatial white noise, and the larger the white noise gain is, the stronger the noise reduction capability is.
Calculating the frequency point k according to the following formulaaUpper reference weight coefficient
Figure BDA0003177929780000113
And kbUpper reference weight coefficient
Figure BDA0003177929780000114
White noise gain decibel value of corresponding super-directional beam forming
Figure BDA0003177929780000115
And
Figure BDA0003177929780000116
Figure BDA0003177929780000117
in the above formula, WSDB,dl(k) The white noise gain formed by the super-directional beam corresponding to the reference weight coefficient at the frequency point k is represented and obtained according to the following formula:
Figure BDA0003177929780000118
wherein H is a conjugate transpose flag, and the calculation of the parameters related to equation (3) may specifically refer to equation (1), which is not described herein again.
Step S231, respectively calculating frequency point kaAnd kbUpper diagonal loading factor initial value.
According to the frequency point k obtained in the step S230aAnd kbWhite noise gain decibel value formed by super-directional beam corresponding to reference weight coefficient
Figure RE-GDA0003311890550000121
And
Figure RE-GDA0003311890550000122
Figure RE-GDA0003311890550000123
obtaining the frequency point k by the following formulaaAnd kbInitial value of diagonal loading coefficient of
Figure RE-GDA0003311890550000124
And
Figure RE-GDA0003311890550000125
Figure BDA0003177929780000126
in the above formula, α1And alpha2Representing the reference value of the diagonal loading coefficient,
Figure BDA0003177929780000127
representing the target white noise gain floor decibel value.
Step S24, adjusting the diagonal loading coefficient reference value, and circularly iterating the steps S22 and S23 until the frequency point k obtained by the last iteration calculationaAnd kbThe difference between the white noise gain decibel value corresponding to the diagonal loading coefficient and the target white noise gain base limit decibel value is smaller than the minimum error threshold, the iteration process is ended, and the frequency point k is obtainedaAnd kbUpper diagonal loading factor alpha (k)a) And alpha (k)b)。
Repeating the steps S22 and S23 t times (the initial value is 0), and calculating the frequency point k after the t iterationaAnd kbDiagonal loading factor of
Figure BDA0003177929780000128
And
Figure BDA0003177929780000129
corresponding white noise gain decibel value
Figure BDA00031779297800001210
And
Figure BDA00031779297800001211
gamma is a stepping factor, and the iterative calculation of the diagonal loading coefficient comprises the following steps:
Figure BDA00031779297800001212
Figure BDA00031779297800001213
when it is satisfied with
Figure BDA00031779297800001214
Then, the optimal solution is approached, the iterative process is ended, and the frequency point k is obtainedaAnd kbDiagonal loading factor of
Figure BDA00031779297800001215
Figure BDA00031779297800001216
Otherwise, the iterative calculation process of step S24 is repeated.
Step S25, according to the diagonal loading coefficient alpha (k)a) And alpha (k)b) Calculating a variable diagonal loading coefficient alphaVSDB(k) (as shown in FIG. 3):
Figure BDA0003177929780000131
and step S26, calculating the weight coefficient of the variable diagonal loading super-directional beam forming at all frequency points.
As shown in fig. 3, according to formula (1), the weight coefficient w for variable diagonal loading super-directional beam forming at all frequency points k is calculatedVSDB(k) Specifically, the following are shown:
Figure BDA0003177929780000132
step S3, according to the requirement of beam width or side lobe suppression degree, calculating the weight coefficient of plane tangential ratio Schonflow beam forming, and determining the directional factor value range of constant beam width beam forming.
The method comprises the following substeps:
and step S31, calculating the Chebyshev coefficients of the dimensions in the directions of the x axis and the y axis respectively, and establishing a Chebyshev coefficient vector of the rectangular plane array.
Chebyshev coefficient h of x-axis direction dimension(x)＝[GH]-1e1Wherein the beam zero position steering vector matrix G ═ G (0), G (ψ)1)，g(ψ2)…g(ψN-1)]Column vector in matrix
Figure RE-GDA0003311890550000133
ψpFor the positions of nulls of the chebyshev weighted beam pattern,
Figure RE-GDA0003311890550000134
(N is the number of columns of the rectangular array in the x-axis direction), and the variable
Figure RE-GDA0003311890550000135
Vector e is determined by side lobe suppression degree R1＝[1，0，...，0]T。
Chebyshev coefficient h of y-axis direction dimension(y)＝[GH]-1e1Wherein the beam zero position steering vector matrix G ═ G (0), G (ψ)1)，g(ψ2)…g(ψM-1)]Column vector in matrix
Figure RE-GDA0003311890550000141
ψpFor the positions of nulls of the chebyshev weighted beam pattern,
Figure RE-GDA0003311890550000142
(M is the number of rows in the y-axis direction of the rectangular array), and the variable
Figure RE-GDA0003311890550000143
Vector e is determined by side lobe suppression degree R1＝[1，0，...，0]T。
The L-dimensional Chebyshev coefficient vector h for a regular rectangular planar array can be represented by h(x)And h(y)Establishing, vector elements
Figure RE-GDA0003311890550000144
And step S32, calculating a weight coefficient formed by the planar Chebyshev beam.
As shown in fig. 3, the chebyshev beamformer is computed at frequency f according to the following equationkWeight coefficient w ofCHEB(k) Comprises the following steps:
wCHEB(k)＝diag{h}v＝ΛLv(k) (9)
in the above formula, ΛLLet diag { h } denote a diagonal matrix with the L-dimensional chebyshev coefficient vector h as the matrix diagonal elements. v (k) represents a guide vector formed by the direction of the target sound source, as explained in equation (1). Chebyshev beamforming is used to effectively control the main lobe width or the side lobe suppression degree while not significantly losing directivity and noise suppression.
Step S33, calculating directivity factor decibel values formed by Chebyshev beams and variable diagonal loading super-directional beams at all frequency points, and respectively selecting a maximum value from the directivity factor decibel values formed by the two beams to form a directivity factor value interval.
And calculating a directivity factor D (k) and evaluating the suppression capability of the directivity noise, wherein the physical quantity represents the array gain under the condition of the isotropy noise. The directivity factor and the white noise gain are mutually exclusive, and the maximum possible directivity factor on the basis of ensuring the white noise floor is required.
The directivity factor decibel value of Chebyshev beam formation at all frequency points K (K is more than or equal to 0 and less than or equal to K) is calculated according to the following formula.
Figure BDA0003177929780000151
In the above formula, DCHEB(k) The directivity factor of Chebyshev beam forming at all frequency points K (K is more than or equal to 0 and less than or equal to K) is obtained according to the following formula:
Figure BDA0003177929780000152
in the above formula, h represents an L-dimensional Chebyshev coefficient vector, ΛLRepresenting a diagonal matrix, Γ, having as matrix diagonal elements the L-dimensional Chebyshev coefficient vector hd(k) Representing the pseudo-coherent matrix of an L x L dimensional microphone array in a diffuse noise field, H being a conjugate transpose sign, v (k) representing the meshAnd the direction of the phonetic transcription source forms a guide vector.
And calculating the directivity factor decibel value of the variable diagonal loading super-directional beam forming at all frequency points K (K is more than or equal to 0 and less than or equal to K) according to the following formula.
Figure BDA0003177929780000153
In the above formula, DVSDB(k) The directivity factor of variable diagonal loading super-directional beam forming at all frequency points K (K is more than or equal to 0 and less than or equal to K) is obtained according to the following formula:
Figure BDA0003177929780000154
in the above formula, v (k) represents a guide vector formed by the direction of the target sound source, Γα(k) Denotes a diagonally loaded autocorrelation matrix, Γ, at frequency point kd(k) Representing the L x L dimension microphone array pseudo-coherence matrix in a diffuse noise field, H is the conjugate transpose flag.
Selecting the maximum directivity factor formed by Chebyshev beam on all frequency points
Figure RE-GDA0003311890550000155
The frequency corresponding to the maximum is expressed as
Figure RE-GDA0003311890550000156
Selecting the maximum value of the directivity factor formed by the variable diagonal loading wave beams on all frequency points
Figure RE-GDA0003311890550000157
The maximum value corresponds to a frequency of
Figure RE-GDA0003311890550000161
The directivity factor value interval is formed by the maximum directivity factor formed by Chebyshev beam and the maximum directivity factor formed by variable diagonal loading beam
Figure RE-GDA0003311890550000162
As shown in fig. 3.
Step S4, determining a constant directivity factor value according to the requirements for white noise suppression and directional noise suppression, dividing the frequency range into a plurality of frequency intervals, and calculating wideband beam weight coefficients optimized for the frequency intervals, respectively, to update the weight coefficients.
The method comprises the following substeps:
step S41, determining the directivity factor according to the requirements for white noise and directivity noise suppression, and dividing the frequency interval based on the number of the intersection points of the directivity factor and the directivity factor curve formed by the variable diagonal loading super-directional beam and the corresponding frequency point positions.
Determining the directivity factor D according to the requirements for white noise and directional noise suppressionc∈ΩD(interval of directivity factor value), D is judgedcDirectivity factor curve D with variable diagonally loaded super-directional beams varying in frequencyVSDBThe number of intersections and the positions of the corresponding frequency points are divided into the following cases (as shown in fig. 3):
(a) if D iscAnd DVSDBWithout intersections, i.e. satisfying D at any frequencyc＜DVSDB(k) Then, the frequency band is not divided, and the whole frequency domain is defined as a constant beamwidth frequency interval.
(b) If D iscAnd DVSDBWith and only one intersection point, corresponding to frequency fcrossAnd satisfy
Figure BDA0003177929780000163
Divided into two frequency bands, constant beam width frequency band interval [0, fcross]And a high frequency band region (f)cross,fs/2]。
(c) If D iscAnd DVSDBWith and only one intersection point, corresponding to frequency fcrossAnd satisfy
Figure BDA0003177929780000164
Is divided intoTwo frequency bands, constant beam width band interval [ f ]cross,fs/2]And low frequency band interval [0, fcross)。
(d) If D iscAnd DVSDBHaving two intersections corresponding to the frequency fcross1And fcross2And satisfy fcross1＜fcross2Is divided into three frequency bands, the low frequency band is divided into [0, fcross1) Constant beam wide band interval fcross1,fcross2]And a high frequency band interval (f)cross2,fs/2]。
And step S42, controlling the weight coefficient formed by the broadband wave beam by utilizing the array gain adjustment parameter in a sub-band manner, and updating the weight coefficient.
If in the constant beam wide frequency band interval, the weight coefficient formed by the broadband beam is controlled by the array gain adjusting parameter beta (k) because the directivity factor is constant DHB(k)＝DcAnd the array gain adjusting parameter beta (k) is a real number, and the array gain adjusting parameter beta (k) can be obtained by solving a one-dimensional quadratic equation set Abeta2(k)+Bβ2(k) A solution of + C ═ 0, where
Figure BDA0003177929780000171
Solving the equation to obtain
Figure BDA0003177929780000172
And
Figure BDA0003177929780000173
β-(k) the white noise gain obtained is larger, so the array gain is adjusted by the parameter
Figure BDA0003177929780000174
As shown in fig. 3, the weight coefficients at the frequency points corresponding to the constant beamwidth interval are calculated according to the following formula.
Figure BDA0003177929780000175
As shown in fig. 3, if β (k) is selected to be 0 in the low frequency band interval, which is equivalent to using variable diagonal loading super-directional beam forming weight coefficients, the weight coefficients at the corresponding frequency points in the low frequency band interval are calculated in step S2.
As shown in fig. 3, if β (k) ∞isselected in the high-band segment, which is equivalent to the weight coefficient formed by using the tangential-ratio schmidt beam, the weight coefficient at the corresponding frequency point in the high-band segment is calculated in step S3.
And updating the weight coefficient.
Updating the calculated wideband beam weight coefficient optimized by frequency division, which is specifically expressed as:
Figure BDA0003177929780000181
and S5, performing beam forming calculation on the frequency domain signal obtained currently in the step S1 by adopting the updated broadband beam weight coefficient, obtaining a time domain signal through inverse Fourier transform, repeatedly executing the step S5 if the incoming wave direction of the target sound source pointed by the beam is not changed, otherwise executing the steps S1-S4, calculating a new weight coefficient and updating the new weight coefficient for the step S5.
As shown in fig. 3, the beam forming calculation y (k) ═ w is performed on the frequency domain signal x (k) of each frame array obtained in step S1 by using the weight coefficient w (k) updated in step S4H(k) And X (k), outputting the frequency domain signal Y (k), and restoring the time domain signal after inverse Fourier transform.
If the incoming wave direction of the target sound source pointed by the wave beam is not changed, the weight coefficient is not required to be updated in real time by depending on signals, and step S5 is repeatedly executed to output a time domain signal frame; if the change has occurred, the real-time signal processing simultaneously performs steps S1-S4 in another process, calculates new weight coefficients and updates for step S5.
Fig. 4 shows a white noise gain comparison of the present invention with a conventional beamforming method. Obviously, the delay-sum beam forming and Chebyshev beam forming method can obtain larger and constant white noise gain in a frequency domain, the super-directional beam forming white noise gain is smaller than the former two white noise gains and is exponentially reduced along with the reduction of the frequency, so that the serious low-frequency noise amplification is caused to cause the method to be invalid, the fixed coefficient diagonal loading method improves the white noise gain background noise, but the white noise gain is about-15 dB in a medium and low frequency range, and the noise reduction performance is influenced. The invention improves the white noise gain in the whole frequency domain, particularly, the white noise gain in a medium and low frequency range can be improved to-5-0 dB by selecting proper parameters, and the beam former of the invention is ensured not to amplify low frequency white noise obviously.
Fig. 5 shows a directional factor comparison of the present invention with a conventional beamforming method. The super-directional beam forming method has the best directivity, and the directivity factor values of delay sum beam forming and Chebyshev beam forming are lower, the invention obtains the directivity factor value index meeting the practical application requirement on the full frequency band through the sub-band weight value optimization technology, obtains the constant directivity on most frequency ranges, for example, the directivity of the frequency part above 1300Hz is constantly 14dB in fig. 5, and realizes the constant beam width while meeting the directivity requirement. Therefore, the invention utilizes the sub-band weight optimization strategy and combines the beam forming method to calculate the optimal frequency variable diagonal loading coefficient, and white noise is not obviously amplified while better beam directivity is obtained.
Fig. 6 is a comparison of the frequency and angle dependent beam response plots for the delay-and-sum beamforming method of the present invention. Fig. 6a is a beam response diagram of a delay-and-sum beamforming method, and 6b and 6c are directivity factors D, respectivelyc13 and Dc18, the beam response diagram of the beamforming method of the present invention. It can be seen that the directivity of the invention is obviously superior to that of the delay-sum beam forming method at medium and low frequencies, the beam width is narrower and the out-of-band rejection is higher, and meanwhile, the beam width can be regarded as constant at most frequencies, thereby reducing signal distortion and ensuring the robustness of the beam forming method. DcThe main lobe width of the constant beam becomes narrower at 18 hours, the directivity is better, but accordinglyBelow 1700Hz the beam begins to widen gradually, while DcWhen 13, the constant beam main lobe width is wider, but the constant beam can be kept in a wider frequency range, and the parameter balance performance index can be flexibly set according to actual requirements.
Fig. 7 is a comparison of the output signal waveforms of the method of the present invention and a conventional beamforming method. Fig. 7a is an input data waveform, which contains a target signal, other directional interference signals and noise, fig. 7b is a delay-sum beam forming method, which outputs a waveform that obviously reduces white noise but does not obviously suppress the interference signals, fig. 7c is a super-directional beam forming method, which outputs a waveform that greatly suppresses the interference noise but also obviously amplifies the white noise, and fig. 7d is a method which can suppress directional interference, and simultaneously, solves the problem of amplifying medium and low frequency white noise, and effectively improves beam forming directivity and output signal-to-noise ratio.
In addition, as shown in fig. 8, an embodiment of the present invention further provides a device for wideband beamforming with optimized subband weights in a planar array, which includes a processor 32 and a memory 31, and may further include a communication component, a sensor component, a power supply component, a multimedia component, and an input/output interface according to actual needs. The memory, communication components, sensor components, power components, multimedia components, and input/output interfaces are all connected to the processor 32. As mentioned above, the memory 31 may be a Static Random Access Memory (SRAM), an Electrically Erasable Programmable Read Only Memory (EEPROM), an Erasable Programmable Read Only Memory (EPROM), a Programmable Read Only Memory (PROM), a Read Only Memory (ROM), a magnetic memory, a flash memory, etc.; the processor 32 may be a Central Processing Unit (CPU), Graphics Processing Unit (GPU), field programmable logic gate array (FPGA), Application Specific Integrated Circuit (ASIC), Digital Signal Processing (DSP) chip, or the like. Other communication components, sensor components, power supply components, multimedia components, etc. may all be implemented using common components in existing smartphones, and are not specifically described herein.
In addition, the wideband beam forming apparatus with plane array and frequency division weight optimization provided in the foregoing embodiment includes a processor 32 and a memory 31, where the processor 32 reads a computer program or an instruction in the memory 31 to perform the following operations:
and step S1, obtaining the spatial position of the planar microphone array and the incoming wave direction information of the target sound source, collecting the environmental sound signals in real time by using the planar microphone array, and converting the environmental sound signals into corresponding frequency domain signals.
And step S2, calculating the variable diagonal loading coefficient and the corresponding weight coefficient for forming the super-directional beam according to the requirement on noise suppression.
Step S3, according to the requirement of beam width or side lobe suppression degree, calculating the weight coefficient of plane tangential ratio Schonflow beam forming, and determining the directional factor value range of constant beam width beam forming.
Step S4, determining a constant directivity factor value according to the requirements for white noise suppression and directional noise suppression, dividing the frequency range into a plurality of frequency intervals, and calculating wideband beam weight coefficients optimized for the frequency intervals, respectively, to update the weight coefficients.
And S5, performing beam forming calculation on the frequency domain signal obtained currently in the step S1 by adopting the updated broadband beam weight coefficient, obtaining a time domain signal through inverse Fourier transform, repeatedly executing the step S5 if the incoming wave direction of the target sound source pointed by the beam is not changed, otherwise executing the steps S1-S4, calculating a new weight coefficient and updating the new weight coefficient for the step S5.
In addition, an embodiment of the present invention further provides a computer-readable storage medium, where instructions are stored on the computer-readable storage medium, and when the instructions are executed on a computer, the computer is enabled to execute the method for forming wideband beams by using planar array sub-band weight optimization as described in fig. 1, and details of a specific implementation manner of the method are not described herein again.
In addition, an embodiment of the present invention further provides a computer program product including instructions, which when run on a computer, causes the computer to execute the method for forming wideband beams by using planar array with frequency division weight optimization as described in fig. 1, and details of a specific implementation manner thereof are not repeated herein.
Compared with the prior art, the method and the device for forming the wideband beam by optimizing the planar array sub-band weight have the following characteristics:
(1) by utilizing a sub-band weight coefficient optimization strategy, frequency intervals are automatically divided according to the requirements on directivity and noise reduction in practical application, and the white noise and directivity interference noise suppression capability of the microphone array and the signal output signal-to-noise ratio are effectively improved.
(2) By calculating the variable diagonal loading coefficient on the frequency, the array gain performance on different frequency points is obviously improved, the problem of low-frequency white noise amplification is solved, and the practical value is high.
(3) The advantages of Chebyshev beam forming and super-directional beam forming are fused, the beam forming side lobe level is better controlled, the constant beam wide beam directivity on a good wide frequency range is achieved, and the effectiveness and the robustness of the beam forming method are improved.
(4) For the planar microphone array with a large number of channels, the requirement on the real-time performance of the operation is obviously reduced, and the design is simple and easy to apply and realize.
The method and apparatus for forming wideband beams by using planar array with frequency-division weight optimization provided by the present invention are described in detail above. It will be apparent to those skilled in the art that various modifications can be made without departing from the spirit of the invention.

Claims (12)

1. A method for forming wideband beam by optimizing planar array sub-band weight is characterized by comprising the following steps:
step S1, obtaining the spatial position of the plane microphone array and the incoming wave direction information of the target sound source, collecting the environmental sound signal in real time by using the plane microphone array, and converting the environmental sound signal into a corresponding frequency domain signal;
step S2, calculating a variable diagonal loading coefficient and a weight coefficient formed by a corresponding super-directional beam according to the requirement on noise suppression;
step S3, calculating a weight coefficient formed by the planar Chebyshev beam according to the requirement on the beam width or the side lobe suppression degree, and determining a directional factor value interval formed by the constant beam width beam;
step S4, determining a constant directivity factor value according to the requirements for white noise suppression and directivity noise suppression, dividing the frequency range into a plurality of frequency intervals, and respectively calculating wideband beam weight coefficients optimized by frequency division to update the weight coefficients;
and S5, performing beam forming calculation on the frequency domain signal obtained currently in the step S1 by adopting the updated broadband beam weight coefficient, obtaining a time domain signal through inverse Fourier transform, repeatedly executing the step S5 if the incoming wave direction of the target sound source pointed by the beam is not changed, otherwise executing the steps S1-S4, calculating a new weight coefficient and updating the new weight coefficient for the step S5.
2. The method of claim 1, wherein the step S2 includes the following sub-steps:
step S21, setting white noise gain bottom limit parameters according to the noise suppression requirements of practical application, and selecting a middle and low frequency interval [ fa,fb]Determining the corresponding frequency point sequence number kaAnd kb；
Step S22, selecting two diagonal loading coefficient reference values, and respectively calculating frequency point kaAnd kbA reference weight coefficient of (c);
step S23, respectively calculating frequency point kaAnd kbInitial value of the diagonal loading coefficient;
step S24, adjusting the diagonal loading coefficient reference value, and circularly iterating the steps S22 and S23 until the frequency point k obtained by the last iteration calculationaAnd kbThe difference between the white noise gain decibel value corresponding to the diagonal loading coefficient and the target white noise gain base decibel value is smaller than the minimum error threshold, the iteration process is ended, and the frequency point k is obtainedaAnd kbUpper diagonal loading factor alpha (k)a) And alpha (k)b)；
Step S25, according to the diagonal loading coefficient alpha (k)a) And alpha (k)b) CalculatingVariable diagonal loading factor alphaVSDB(k)：
Figure FDA0003177929770000021
And step S26, calculating the weight coefficients of the variable diagonal loading super-directional beam forming at all frequency points.
3. The method of claim 2, wherein in steps S22 and S26,
calculating a reference weight coefficient at each frequency point according to the following formula;
Figure FDA0003177929770000022
in the above formula, gammad(k)+α(k)IL＝Γα(k) Denotes a diagonally loaded pseudo-coherent matrix at frequency point k, α (k) denotes a diagonally loaded coefficient, ILIs an identity matrix; gamma-shapedd(k) Representing a pseudo-coherent matrix of an L x L dimensional microphone array in a diffuse noise field, the Zeta column of the L-th row in the matrix having an element [ gamma ]d(k)]l,ζ＝sin(2πfkdl,ζ/c)/(2πfkdl,ζAnd/c) (L-0, 1, L-1, ζ -0, 1, L-1), wherein the equivalent distance in the incoming wave direction between any two microphones is equal to the equivalent distance in the incoming wave direction
Figure FDA0003177929770000023
(ml,nl) And (m)ζ,nζ) Two-dimensional coordinates corresponding to the microphone numbers l and ζ are respectively expressed, θ represents a pitch angle from the direction of the plane wave of the target sound source to the coordinates,
Figure FDA0003177929770000024
representing the horizontal angle from the direction of the plane wave of the target sound source to the coordinate;
Figure FDA0003177929770000025
a steering vector formed by representing the direction of the target sound source,
Figure FDA0003177929770000031
wave numbers of incoming waves of a target sound source in an x axis and a y axis respectively, c is sound propagation speed in the environment, (x is the sound propagation speed of the target sound source in the environmentl,yl) Representing two-dimensional coordinates of the target audio source.
4. The method of claim 3, wherein the step S23 includes the following sub-steps:
step S230, respectively calculating frequency point kaAnd kbWhite noise gain decibel value formed by the super-directional beam corresponding to the reference weight coefficient;
step S231, respectively calculating frequency point kaAnd kbUpper diagonal loading factor initial value.
5. The method of claim 4, wherein in step S230:
calculating the frequency point k according to the following formulaaAnd kbWhite noise gain decibel value formed by the super-directional beam corresponding to the reference weight coefficient;
Figure FDA0003177929770000032
in the above formula, WSDB,dl(k) The white noise gain formed by the super-directional beam corresponding to the reference weight coefficient at the frequency point k is represented and obtained according to the following formula:
Figure FDA0003177929770000033
where H is the conjugate transpose flag.
6. The tablet of claim 5The method for forming wideband beam by optimizing area array sub-band weight is characterized in that in step S231, frequency point k is obtained according to the following formulaaAnd kbInitial value of the diagonal loading coefficient;
Figure FDA0003177929770000034
in the above formula, α1And alpha2Representing the reference value of the diagonal loading coefficient,
Figure FDA0003177929770000035
and expressing the target white noise gain floor decibel value.
7. The method of claim 1, wherein the step S3 includes the following sub-steps:
step S31, calculating the Chebyshev coefficients of the dimensions in the x and y directions respectively, and establishing the Chebyshev coefficient vector of the rectangular plane array
Step S32, calculating a weight coefficient formed by the planar Chebyshev beam;
step S33, calculating directivity factor decibel values formed by Chebyshev beams and variable diagonal loading super-directional beams at all frequency points, and respectively selecting a maximum value from the directivity factor decibel values formed by the two beams to form a directivity factor value interval.
8. The method of claim 7, wherein in step S32, the Chebyshev beamformer is computed at frequency f according to the following formulakWeight coefficient w ofCHEB(k) Comprises the following steps:
wCHEB(k)＝diag{h}v＝ΛLv(k)
in the above formula, ΛLD, representing a diagonal matrix with the L-dimensional Chebyshev coefficient vector h as a matrix diagonal element; v (k) a director formed by the direction of the target sound sourceAmount of the compound (A).
9. The method according to claim 7, wherein in step S33, the directivity factor decibel values of chebyshev beamforming at all frequency points K (K is greater than or equal to 0 and less than or equal to K) are calculated according to the following formula;
Figure FDA0003177929770000041
in the above formula, DCHEB(k) The directivity factor of Chebyshev beam forming at all frequency points K (K is more than or equal to 0 and less than or equal to K) is obtained according to the following formula:
Figure FDA0003177929770000042
in the above formula, h represents an L-dimensional Chebyshev coefficient vector, ΛLRepresenting a diagonal matrix, Γ, having as matrix diagonal elements the L-dimensional Chebyshev coefficient vector hd(k) Representing a pseudo-coherent matrix of an L multiplied by L dimensional microphone array in a diffused noise field, H is a conjugate transpose mark, and v (k) represents a steering vector formed by the direction of a target sound source;
calculating the directivity factor decibel value of variable diagonal loading super-directional beam forming at all frequency points K (K is more than or equal to 0 and less than or equal to K) according to the following formula;
Figure FDA0003177929770000051
in the above formula, DVSDB(k) Expressing the directivity factor formed by variable diagonal loading super-directional beam at all frequency points K (K is more than or equal to 0 and less than or equal to K), and obtaining the directivity factor according to the following formula;
Figure FDA0003177929770000052
in the above formulaV (k) a guide vector formed by the direction of the target sound source, Γα(k) Representing a diagonally loaded pseudo-coherence matrix at frequency point k.
10. The method of claim 1, wherein the step S4 includes the following sub-steps:
step S41, determining a directivity factor according to the requirements for white noise and directivity noise suppression, and dividing a frequency interval based on the number of intersections of the determined directivity factor and a directivity factor curve formed by variable diagonal loading super-directional beam and the corresponding frequency point positions;
and step S42, controlling the weight coefficient formed by the broadband wave beam by utilizing the array gain adjustment parameter in a sub-band manner, and updating the weight coefficient.
11. The method of claim 1, wherein the method comprises:
calculating a weight coefficient at a frequency point corresponding to a constant beam width frequency band interval according to the following formula;
Figure FDA0003177929770000053
in the above formula, gammaα(k) Represents a diagonally loaded pseudo-coherent matrix at a frequency point k, beta (k) represents a gain adjustment parameter of the matrix, and ΛLA diagonal matrix in which L-dimensional chebyshev coefficient vector h is a matrix diagonal element, and v (k) a steering vector formed by the directions of the target sound sources.
12. A flat array, band-division weight optimized wideband beamforming apparatus comprising a processor and a memory, said processor reading a computer program or instructions in said memory for performing the following operations:
step S1, obtaining the spatial position of the plane microphone array and the incoming wave direction information of the target sound source, collecting the environmental sound signal in real time by using the plane microphone array, and converting the environmental sound signal into a corresponding frequency domain signal;
step S2, calculating a variable diagonal loading coefficient and a weight coefficient formed by a corresponding super-directional beam according to the requirement on noise suppression;
step S3, calculating a weight coefficient formed by the planar Chebyshev beam according to the requirement on the beam width or the side lobe suppression degree, and determining a directional factor value interval formed by the constant beam width beam;
step S4, determining a constant directivity factor value according to the requirements for white noise suppression and directivity noise suppression, dividing the frequency range into a plurality of frequency intervals, and respectively calculating wideband beam weight coefficients optimized by frequency division to update the weight coefficients;
and S5, performing beam forming calculation on the frequency domain signal obtained currently in the step S1 by adopting the updated broadband beam weight coefficient, obtaining a time domain signal through inverse Fourier transform, repeatedly executing the step S5 if the incoming wave direction of the target sound source pointed by the beam is not changed, otherwise executing the steps S1-S4, calculating a new weight coefficient and updating the new weight coefficient for the step S5.
CN202110838291.1A 2021-07-23 2021-07-23 Planar array frequency division band weight optimization broadband beam forming method and device Active CN113702900B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110838291.1A CN113702900B (en) 2021-07-23 2021-07-23 Planar array frequency division band weight optimization broadband beam forming method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110838291.1A CN113702900B (en) 2021-07-23 2021-07-23 Planar array frequency division band weight optimization broadband beam forming method and device

Publications (2)

Publication Number Publication Date
CN113702900A true CN113702900A (en) 2021-11-26
CN113702900B CN113702900B (en) 2024-03-08

Family

ID=78650359

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110838291.1A Active CN113702900B (en) 2021-07-23 2021-07-23 Planar array frequency division band weight optimization broadband beam forming method and device

Country Status (1)

Country Link
CN (1) CN113702900B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102082591A (en) * 2010-12-13 2011-06-01 武汉滨湖电子有限责任公司 Method for forming circular array antenna digital wave beams
CN104993861A (en) * 2015-05-21 2015-10-21 中国电子科技集团公司第十研究所 Array element selectivity nulling antenna beam sysnthesis method
CN110174656A (en) * 2019-05-21 2019-08-27 电子科技大学 A kind of thinned array design method and device based on frequency domain broad-band EDFA
CN111208213A (en) * 2020-02-25 2020-05-29 重庆大学 Spectral seeking sub-band minimum variance ultrasonic imaging algorithm fused with alternative multiplier iteration

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102082591A (en) * 2010-12-13 2011-06-01 武汉滨湖电子有限责任公司 Method for forming circular array antenna digital wave beams
CN104993861A (en) * 2015-05-21 2015-10-21 中国电子科技集团公司第十研究所 Array element selectivity nulling antenna beam sysnthesis method
CN110174656A (en) * 2019-05-21 2019-08-27 电子科技大学 A kind of thinned array design method and device based on frequency domain broad-band EDFA
CN111208213A (en) * 2020-02-25 2020-05-29 重庆大学 Spectral seeking sub-band minimum variance ultrasonic imaging algorithm fused with alternative multiplier iteration

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
任超;吴嗣亮;王菊;李加琪;: "基于可变对角加载的稳健空时波束形成算法", 北京理工大学学报, no. 11 *
王之海;王大成;曾武;: "利用Chebyshev窗函数获得恒定束宽加权矩阵的数值算法", 海洋技术, no. 03 *

Also Published As

Publication number Publication date
CN113702900B (en) 2024-03-08

Similar Documents

Publication Publication Date Title
US9591404B1 (en) Beamformer design using constrained convex optimization in three-dimensional space
US9031257B2 (en) Processing signals
CN105355210B (en) Preprocessing method and device for far-field speech recognition
EP1429581B1 (en) Method of broadband constant directivity beamforming for non linear and non axi-symmetric sensor arrays embedded in a obstacle
JP2021500634A (en) Target voice acquisition method and device based on microphone array
CN113655436B (en) Method and device for optimizing broadband wave beam formation through channel calibration particle swarm
CN113422630B (en) Self-adaptive focusing broadband beam forming method and system
JP2017531971A (en) Calculation of FIR filter coefficients for beamforming filters
CN108877828B (en) Speech enhancement method/system, computer-readable storage medium, and electronic device
CN107577872B (en) Time domain frequency invariant beam former design method and device
CN112735461A (en) Sound pickup method, related device and equipment
KR102514060B1 (en) A method of beamforming sound for driver units in a beamforming array and sound apparatus
CN109040913B (en) Beam forming method of window function weighting electroacoustic transducer transmitting array
Cho et al. Adaptive near-field beamforming techniques for sound source imaging
CN113702900A (en) Planar array sub-band weight optimization broadband beam forming method and device
CN110927668A (en) Sound source positioning optimization method of cube microphone array based on particle swarm
Patel et al. On the design of optimal linear microphone array geometries
JP4292327B2 (en) Speaker array and microphone array
CN116611223B (en) Accurate array response control method and device combined with white noise gain constraint
US20220030353A1 (en) Flexible differential microphone arrays with fractional order
CN114915875B (en) Adjustable beam forming method, electronic equipment and storage medium
CN115086836B (en) Beam forming method, system and beam former
JP7181738B2 (en) Speaker device, speaker coefficient determination device, and program
CN116017230A (en) Microphone array, signal processing method, device, equipment and medium thereof
CN114023307B (en) Sound signal processing method, speech recognition method, electronic device, and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant