CN109001680A - The sparse optimization algorithm of block in auditory localization - Google Patents
The sparse optimization algorithm of block in auditory localization Download PDFInfo
- Publication number
- CN109001680A CN109001680A CN201810660554.2A CN201810660554A CN109001680A CN 109001680 A CN109001680 A CN 109001680A CN 201810660554 A CN201810660554 A CN 201810660554A CN 109001680 A CN109001680 A CN 109001680A
- Authority
- CN
- China
- Prior art keywords
- block
- signal
- microphone
- sparse
- sound source
- 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.)
- Pending
Links
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
- G01S5/00—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations
- G01S5/18—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations using ultrasonic, sonic, or infrasonic waves
- G01S5/20—Position of source determined by a plurality of spaced direction-finders
Abstract
The sparse optimization algorithm of block in auditory localization, belongs to field of signal processing, and in order to optimize block-sparse signal, main points are step3: regularization: in above-mentioned index set LjtIn corresponding related coefficient μ, the index L of block of its energy between ceiling capacity and the half of ceiling capacity is picked out againi, i.e. LiMeet:
Description
Technical field
The invention belongs to field of signal processing, further relate to distributed in a kind of compressed sensing based reverberation room
Owe the more sound localization methods of fixed number amount microphone array and system.
Background technique
With the rapid development of the communications industry, array signal processing becomes a research hotspot.Since the 1990s,
Auditory localization technology based on microphone array obtains significant progress and development in every field.Such as in military field,
Auditory localization technology can be used for determining sniper position and estimation Canon launching position etc.；In industrial application, using Mike
Wind array sound source location technology positions the failure of mechanical equipment；In terms of security monitoring, for existing in video monitoring
Dead angle and the bad problem of vision, auditory localization technology can play the role of compensating well for；In audio/video meeting, sound source
The application of location technology and speech enhancement technique makes participant that can freely walk about in the room without manual switching voice collecting
Equipment.However, in actual room environment, sound-source signal is more through room wall, ceiling, ground etc. as meeting room
Secondary reflection and form RMR room reverb.Under the conditions of reverberation, conventional microphone array auditory localization algorithm is such as based on high-resolution
The location algorithm of Power estimation or the positioning performance for reaching the algorithm of delay inequality based on wave seem rather out of strength, and are based on maximum work output
Although anti-reverberation ability is preferable for the steerable beam formation algorithm of rate, but in more sound sources, this algorithm need overcome initial value
Sensitive issue, and in order to promote positioning accuracy, it must be arranged with a large amount of microphone array as cost.
General more auditory localization algorithms are substantially divided to two classes: first is that the auditory localization based on traditional multiple signal classification is calculated
Method, second is that based on blind source separating-sodar temporal difference method auditory localization algorithm.As (Song palace a kind of jade a kind of jade is based on microphone array to document
Indoor voice location algorithm study [D] Nanjing Information engineering Univ, 2016.) propose one kind be based on consistent focusing convert most
The microphone array double sound source location algorithm of small square law, with the broadband signal based on consistent focusing transformation least square method
MUSIC algorithm positions indoor near field double sound source.The characteristics of making a general survey of entire microphone array auditory localization algorithm, it is existing
Microphone array auditory localization technology can not still get rid of the limitation of triangulation location, even if using Blind channel isolation technics, can be extensive
Multiple sound source number still can not be more than number of microphone；It owes determine blind source separation method although and can solve sound source number to be greater than microphone
The problem of quantity, but the case where opposite linear instantaneous mixtures can only be handled at present, Convolution Mixture Signals and nonlinear mixing cannot be handled
The case where.So the microphone array system with practical value is mostly massive array at present, i.e., number of microphone is at least super
Cross sound source number.
It after compressive sensing theory proposes, is applied in array signal processing by many scholars, such as document (Zhao little Yan, week
Beautiful jade, Wu Zhen raise compressed sensing based microphone array auditory localization algorithm [J] Southeast China University's journal (natural science edition),
2015,45 (2): 203-207.) a kind of compressed sensing based auditory localization algorithm is proposed, realizing makes under reverberant ambiance
The location algorithm of 1 sound source is positioned with 3 microphones.The algorithm assumes that 3 microphones and speaker are in same level, will
Sound source spatial spreading that may be present is at 36 positions (0 °, 10 ° ..., 350 °), and with the room impulse response of this 36 positions
To construct dictionary, then sound-source signal vector recovered using OMP algorithm, greatest member correspond to spatial position as reality using in vector
The position of border sound source.A kind of document (sound source three-dimensional positioning method using distributed microphone array of Ke Wei, Zhang Ming, Zhang Tiecheng
[J] acoustic journal, 2017 (3): 361-369.) propose a kind of auditory localization using spatial sparsity and compressed sensing principle
Method, this method pass through cosine transform twice first and extract voice signal property, and with the sparse location model of this feature construction, so
Afterwards using online dictionary learning method dynamic adjustment dictionary, the mismatch problems between sparse model and actual signal are overcome, finally
Use a kind of improved smooth l0The sparse restructing algorithm of norm carries out sound source position resolving.
Although both the above algorithm all realizes the auditory localization of microphone array under reverberant ambiance, microphone number is used
Amount is still more than the sound source quantity in space.In the reverberant ambiance of indoor public places large-scale as hall, auditorium,
Realize more auditory localizations and speech Separation, it is necessary to which a large amount of microphone array acquires equipment, this undoubtedly increases the meter of system
Calculate burden and hardware cost.It is today that the voice communication apparatus as similar mobile phone, computer increasingly minimizes, big and bulky
Hardware facility is obviously out-of-date, reach be accurately positioned target under the premise of, we prefer to using hardware device it is more small and exquisite, just
Victory, in microphone array sound source localization system, reducing array quantity is also an active path for reducing array sizes.
Summary of the invention
In order to optimize block-sparse signal, the following technical solutions are proposed by the present invention:
A kind of sparse optimization algorithm of block in auditory localization, steps are as follows:
Input: compression observing matrix Φ；
Microphone observation signal X；
Degree of rarefication k；
Output: reconstruction signalIt is the estimation of block-sparse signal S；
In following steps: t is the number of iterations, rtIndicate the residual error of t iteration, j is the column serial number for compressing observing matrix Φ
Index, LtSet is indexed for the block after t iteration, mean | | indicate the mean value of element absolute value,To be indexed according to block
Set LtThe matrix that the column picked out from Φ are constituted,ForPseudo inverse matrix；
Step1: initialization: residual error r0=X, the number of iterations t=1, block index set
Step2: related coefficient μ, μ (L are askedj)=ΦT[j]rt-1, select and residual error rt-1The indexed set of most matched k block
It closes:
Step3: regularization: in above-mentioned index set LjtIn corresponding related coefficient μ, its energy is picked out again between most
The index L of block between big energy and the half of ceiling capacityi, i.e. LiMeet:
Step4: it updates block and indexes set Lt=Lt-1∪Li；
Step5: the estimation of signal is calculated:And update residual error
Step6: as t >=k, algorithm terminates；Conversely, t=t+1, continues iteration into step2；
Then last time iteration obtainsThe as estimation of block-sparse signal S
The utility model has the advantages that being much smaller than the dimension of signal, common sparse solution side for the dimension of the observation data X of microphone
Method is helpless for solving such equation group, and optimization algorithm of the invention is able to solve the influence of dimension, realize for
The optimization of block-sparse signal.
Detailed description of the invention
Fig. 1 is the basic flow chart that the present invention realizes.
Fig. 2 is location simulation figure when present invention simulation multi-acoustical is located at the different location in room simultaneously；
Wherein: (a) 4 sound sources are located at the room upper right corner, (b) 5 sound sources be distributed in simultaneously the room upper left corner lower left corner and
The upper right corner, (c) 5 sound sources are distributed in the room upper left corner, the upper right corner, 6, the lower right corner (d) sound source while being located at room upper left simultaneously
Angle, (e) 7 microphones are distributed in the room upper right corner and the lower left corner simultaneously, and (f) 8 microphones are distributed among room simultaneously.
Fig. 3 is position success rate histogram of the present invention under the different reverberation time.
Fig. 4 is position success rate histogram of the present invention in different frequencies points.
Fig. 5 is auditory localization success rate line chart of the present invention when sound source number increases.
Fig. 6 is the positioning comparison diagram of the method for the present invention and other conventional methods.
Specific embodiment
In order to keep technology contents of the invention, feature etc. more obvious and easy to understand, realization of the invention is walked below with reference to Fig. 1
Suddenly it explains, briefly, the more sound localization methods of distributed microphone array in reverberation room of the present invention can be by software
It realizes, relates generally to following step:
(1) room-sized is determined first, i.e., obtains the length numerical value in room respectively, while determining microphone array
Position, as unit of rice.
(2) room impulse response is constructed, compression observing matrix is generated.
(3) relationship between more sound-source signals, compression observing matrix and microphone observation data three is by owing constant linear
Shown in equation group, this is the more auditory localization algorithm models of reverberation proposed by the present invention.
(4) microphone observation data are known quantity, calculate compression observing matrix, acquire sound source using the sparse optimization algorithm of block
The estimation signal is integrated in the estimation of signal, and the position of nonzero block is corresponding with space lattice, sound source sum number that you can get it
Mesh.
Above-mentioned technical proposal is explained in detail as follows:
The characteristics of receiving signal according to the more sound-source signals propagation of reverberation and microphone constructs compressed sensing location algorithm；?
Under the premise of knowing room-sized and microphone position, observing matrix is compressed using Image method construct；Using the sparse optimization of block to this
It invents the deficient more auditory localization algorithms of fixed number amount microphone array in the reverberation room proposed to be solved, obtains each sound in space
The position in source and number.Specific step is as follows:
1. constructing the more auditory localization algorithms of compressed sensing reverberation
1a) the more sound source propagation model descriptions of reverberation
Under room reverberation environment, the multi channel time domain propagation model of the unknown more sound sources of sound source number are as follows:
Wherein xmIndicate m-th of microphone received signal, sn(t) n-th of sound-source signal, h are indicatedmn(t) it indicates n-th
Room impulse response between channel between sound source and m-th of microphone, also known as n-th of sound source and m-th of microphone, *
For convolution symbol, M is the quantity of microphone.It is obtained through Short Time Fourier Transform:
1b) reverberation more sound source space-frequencies domain block sparse expression
Considering that K sound source is distributed in same level simultaneously, i.e. the vertical view face in room, this plane is divided into G grid,
Sound source number is much smaller than the quantity of grid, i.e. K < < G (ratio of general G and K are greater than an order of magnitude), and grid is enough
It is close, so that each sound source only occupies a grid.Assuming that sound source position is constant in time observation window.For convenient for derivation, frame
The two-dimensional grid by column is arranged in one-dimensional vector when construction, and (this frame is equally applicable to three-dimensional grid, only in calculation amount
Difference), F frequency point is taken at each grid g, then the frequency domain representation of each Grid Signal is Sg∈CF×1, g ∈ { 1 ..., G },
The space-frequency domain of entire two-dimensional grid signal where multi-acoustical is expressed asThen it is dispersed in net
Sound source on lattice for entire two-dimensional grid, the design feature of block, and K < < G are presented in the expression of S, makes the table of S
Up to also having sparsity, i.e. expression of the signal S in space-frequency domain has block sparsity.If sound source is not present at certain grid, at this
Value of the signal in space-frequency domain be zero, conversely, the signal at this will appear nonzero value in space-frequency domain.Sound source number K is much smaller than
Number of grid G, so the expression in space-frequency domain has block sparsity, therefore also referred to as S is block-sparse signal for S,
Sparse support collection (nonzero block) corresponds to practical sound source, wherein CF×1Indicate the scale space of F dimension, CFG×1Indicate the ruler of FG dimension
Spend space.
1c) the more auditory localization algorithms of compressed sensing
Assuming that having M microphone in room, and M < K, the observation signal that this M microphone receives is expressed asWherein Xm∈CF×1The frequency spectrum of signal, and M < < G are received for m-th of microphone, then X can regard block as
Compression observation of the sparse signal S under the compression of room impulse response, i.e. X=Φ S, Φ represent room sound source via room
The projection on microphone is acted on, also referred to as compression observing matrix.1b as above) it is described, block-sparse signal S is sparse space-frequency domain
Supported collection (nonzero block) corresponds to actual sound source, therefore restoring this block-sparse signal S from the observation data X of microphone array is
The key that more sound sources are positioned in the present invention, to the more auditory localization problems of reverberation of the invention are described as solving as follows
Owe constant linear equation group problem:
The item that above formula owes the constant linear equation group equation left side indicates M microphone received signal in the value of different frequent points.
First item is respectively to choose F in the projection (room impulse response) from each mesh point in room to each microphone on the right of its equation
The compression observing matrix of a frequency point construction, specific configuration process is referring to following 2 construction compression observing matrix parts, equation the right
The frequency domain representation of Section 2 each mesh point signal in the entire plane grid where sound source.The theoretically grid pair of sound source
The frequency point answered has nonzero value, and the corresponding value of all frequency points of the grid of no sound source is zero, i.e., there are one for the nonzero block in sound source and S
One corresponding relationship, such as 1b) as described in, block-sparse signal S is one made of being stacked by the value of frequency point of each grid in room by column
Dimensional vector, therefore by determining that position and number of the nonzero block in S can obtain the position sum number visual estimation meter of sound source in space.
In addition, more sonic location systems of the invention are equally applicable to the case where individual microphone breaks down, only need at this time by
The corresponding item in the equation left side removes in above formula, while row corresponding in observing matrix being removed, and Section 2, which is not done, on the right of equation changes
Become.So, without adjusting array structure, without location model is changed, this hair still can be used when there is microphone to break down
It is bright to realize more auditory localizations.
2. construction compression observing matrix
2a) generate full room impulse response
Assuming that the rectangular closed structure that room is constituted for the multi-panel wall with reflection coefficient, and known room-sized, it uses
Image method (image method) constructs any space point sound source (position V in roomg) reach a certain position microphone (position μm)
Room impulse response, with Green function representation are as follows:
In formulaF indicates frequency, and τ is time delay, μmWithRespectively indicate m-th of microphone position and g-th
The position of r-th of imaginary source of mesh point, ιrIndicate that the reflection coefficient of r secondary reflection, R are total reflection number, α is that decaying is normal
Number, α=1 in spherical surface propagation model, c are the velocity of sound.
2b) construction compression observing matrix
Using 2a) in sound source-microphone projection construction F frequency point projection matrix are as follows:Then all mesh points project to the corresponding compression of m-th of microphone via room and observe
Template are as follows:So template is observed in the corresponding compression of microphone array are as follows: Φ '=
[Φ'1,…,Φ'M]T, observing matrix Φ=[Φ must be compressed after Φ ' is pressed row normalization1,...,ΦM]T。
3. solving sound source position and number using the sparse optimization algorithm of block
Sound-source signal 3a) is restored using the sparse optimization algorithm of block
Such as 1c) as described in, to determine position and the number of sound source, must first pass through solve 1c) in owe constant linear equation group
Obtain the estimation of signal SDimension M × F of the observation data X of equation left side microphone much smaller than signal dimension G × F, it is general
Logical sparse method for solving is helpless for solving such equation group.But such as 1b) as described in, signal S has block sparsity, institute
With we are using regularization block orthogonal matching pursuit (the Regularized Group in the sparse optimization algorithm of block
Orthogonal Matching Pursuit, ReGOMP) algorithm solves in this way with the sparse equation group of block, the algorithm
Specific step is as follows:
Input: compression observing matrix Φ；
Microphone observation signal X；
Degree of rarefication k；
Output: reconstruction signalIt is the estimation of block-sparse signal S；
In following steps: t is the number of iterations, rtIndicate the residual error of t iteration, j is the column serial number for compressing observing matrix Φ
Index, LtSet is indexed for the block after t iteration, mean | | indicate the mean value of element absolute value,To be indexed according to block
Set LtThe matrix that the column picked out from Φ are constituted,ForPseudo inverse matrix.
Step1: initialization: residual error r0=X, the number of iterations t=1, block index set
Step2: related coefficient μ, μ (L are askedj)=ΦT[j]rt-1, select and residual error rtThe indexed set of -1 most matched k block
It closes:
Step3: regularization: in above-mentioned index set LjtIn corresponding related coefficient μ, its energy is picked out again between most
The index L of block between big energy and the half of ceiling capacityi, i.e. LiMeet:
Step4: it updates block and indexes set Lt=Lt-1∪Li；
Step5: the estimation of signal is calculated:And update residual error
Step6: as t >=k, algorithm terminates；Conversely, t=t+1, continues iteration into step2；
Then last time iteration obtainsAs 1c) in block-sparse signal S estimation on the right of equation group in Section 2
3b) determine sound source position and number
Step 3a) estimation of block-sparse signal solved using the sparse optimization algorithm of blockIn order to determine the position of sound source
And number, the signal that must be first G × F to dimensionIn element carry out piecemeal processing.One group of every F element, and ask respectively each
The l of F element of group2Norm, i.e.,G ∈ [1, G], such as 1b) as described in, if there is sound source at grid, then herein
F value of frequency point will appear nonzero value, l2Norm is naturally also nonzero value, conversely, if sound source is not present at grid, this F
The l of value of frequency point2Norm is zero, it is such grouping so thatBlock structure is presented；Such as 1c) as described in,Sparse support collection (non-zero
Block) and the practical sound source in room between there are one-to-one relationship, and S is made of being stacked as room grid by column, then to manage
By upper,According to same sequence remove corresponding grid, the position of nonzero block will correspond to the position of sound source within a grid.According to
After the method piecemeal of front, the nonzero block (l of non-zero2Norm value, i.e.,) number be exactly sound
The number in source.This method can position multi-acoustical simultaneously using the microphone array less than sound source number in reverberation room, lead to
It crosses reduction number of microphone and reduces calculation amount, also reduce the hardware cost of sonic location system, and real-time limited for volume
Very attractive for more demanding voice communication apparatus.
The present invention proposes a kind of more sound localization methods of distributed microphone array in reverberation room, and this method is not by battle array
Type constraint, and can overcome the limitation of triangulation location, and using the microphone array of deficient fixed number amount, (number of microphone is less than sound source
Number) multi-acoustical in reverberation room can be positioned.
In existing more auditory localization algorithms, microphone number cannot be less than sound source number；The increase of number of microphone is not only
Mean the complexity increase of algorithm, hardware cost will also improve, and the present invention uses the microphone array realization pair for owing fixed number amount
It is positioned while multi-acoustical, more meets the trend of voice communication apparatus miniaturization.
Microphone array of the invention is not constrained by formation, can arbitrarily be placed, and distributive array is suitble to；And number of microphone
Increase and decrease does not influence algorithm frame, only need to simply increase and decrease compression observing matrix line number.
Effect of the invention can be further illustrated by following emulation:
Simulated conditions: the present invention is respectively 3.3m in a length, and the rectangle reverberation room of 4.4m, 2.5m use two
A omnidirectional microphone carries out positioning performance verifying.Room two-dimensional surface is divided into 10 × 10 using 0.3m and 0.4m as interval respectively
Grid, two microphone position coordinates are respectively [1.6,1.6,1.7] and [1.4,1.9,1.7], sound source and microphone array
In same level, sample frequency Fs=16kHZ, velocity of sound c=343m/s, wall reflection coefficient is 0.7, using Image
The full room impulse response length that method generates is about 2500 points or so, the unified value for intercepting preceding 1000 frequency points, frequency domain into
The Fourier transformation that 2048 points of row, and F frequency point construction compression observing matrix is chosen at equal intervals.
Emulation content:
Emulation 1, positioning scenarios of the invention are simulated when being located at room different location simultaneously to different sound sources.When reverberation
Between 0.3s, frequency point number F be 60；* is sound source position in Fig. 2 (a-f), and o then represents the sound source position estimated using the present invention,
Δ indicates two microphone positions.
From 4~8 sound sources that can be seen that in Fig. 2 (a-f) for different location in reverberation room, two wheats are used only
Gram wind can position it, Considerable effect.
Emulation 2, reverberation time RT600.2s, 0.3s, 0.4s, 0.5s, 0.6s are taken respectively, take F=40 frequency point structure at equal intervals
Compression observing matrix is made, more auditory localizations, shadow of the test reverberation time to position success rate of the invention are carried out using the present invention
It rings.The histogram of Fig. 3 demonstrates anti-reverberation performance of the invention.
Emulation 4, reverberation time RT60=0.3s, frequency point number F take 20,40,60,80,100,120 respectively, test frequency point number with
The relationship of position success rate of the invention, such as Fig. 4, it is seen that in general, position success rate increases with the increase of frequency point number, but
When exceeding a certain range, this increased trend mitigates gradually.
Emulation 5, reverberation time RT60=0.3s, frequency point number are F=60, and orientable sound source number is respectively 3,4,5,6,7
When position success rate such as Fig. 5 of the invention, it is seen that the case where being 3 and 4 for sound source number, positioning result of the invention is can be with
Receive, because the present invention is used only 2 microphones and carries out more auditory localizations, the case where for more sound sources, the present invention
In fail to obtain good effect, but be enough to illustrate that the present invention carries out the latent of more auditory localizations using the microphone for owing fixed number amount
Power.
Emulation 6, reverberation time RT60=0.3s, frequency point number is F=60, by the present invention to more sound sources under the conditions of reverberation
Positioning scenarios are respectively with the location algorithm based on delay inequality, the location algorithm based on High-Resolution Spectral Estimation and based on maximum defeated
The steerable beam formation algorithm of power compares out, as a result such as Fig. 6.It can be seen that for 3 sound randomly placed in reverberation room
Source, location algorithm of the invention can be accurately positioned out its position, and other three kinds of algorithms are affected by reverberation etc. or other are former
Cause cannot be accurately positioned out sound source position, or even have the case where leakage positions.
In conclusion sonic location system of the invention uses the microphone less than sound source quantity not only to can be carried out more accurately
The more auditory localizations of reverberation；The reduction of number of microphone reduces the calculation amount of algorithm simultaneously, in practical applications hardware cost
Also it reduces therewith；The present invention has the characteristic of distributed microphone, is not constrained by formation, and application is more flexible, and is suitable for individual
The case where microphone breaks down；Compared with other algorithms, anti-reverberation ability is more preferable.
The preferable specific embodiment of the above, only the invention, but the protection scope of the invention is not
It is confined to this, anyone skilled in the art is in the technical scope that the invention discloses, according to the present invention
The technical solution of creation and its inventive concept are subject to equivalent substitution or change, should all cover the invention protection scope it
It is interior.
Claims (1)
1. the sparse optimization algorithm of block in a kind of auditory localization, which is characterized in that steps are as follows:
Input: compression observing matrix Φ；
Microphone observation signal X；
Degree of rarefication k；
Output: reconstruction signalIt is the estimation of block-sparse signal S；
In following steps: t is the number of iterations, rtIndicate the residual error of t iteration, j is the rope for compressing the column serial number of observing matrix Φ
Draw, LtSet is indexed for the block after t iteration, mean | | indicate the mean value of element absolute value,Gather to be indexed according to block
LtThe matrix that the column picked out from Φ are constituted,ForPseudo inverse matrix；
Step1: initialization: residual error r0=X, the number of iterations t=1, block index set
Step2: related coefficient μ, μ (L are askedj)=ΦT[j]rt-1, select and residual error rt-1The index set of most matched k block:
Step3: regularization: in above-mentioned index set LjtIn corresponding related coefficient μ, its energy is picked out again between maximum energy
The index L of block between amount and the half of ceiling capacityi, i.e. LiMeet:
Step4: it updates block and indexes set Lt=Lt-1∪Li；
Step5: the estimation of signal is calculated:And update residual error
Step6: as t >=k, algorithm terminates；Conversely, t=t+1, continues iteration into step2；
Then last time iteration obtainsThe as estimation of block-sparse signal S
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810660554.2A CN109001680A (en) | 2018-06-25 | 2018-06-25 | The sparse optimization algorithm of block in auditory localization |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810660554.2A CN109001680A (en) | 2018-06-25 | 2018-06-25 | The sparse optimization algorithm of block in auditory localization |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109001680A true CN109001680A (en) | 2018-12-14 |
Family
ID=64600561
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810660554.2A Pending CN109001680A (en) | 2018-06-25 | 2018-06-25 | The sparse optimization algorithm of block in auditory localization |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109001680A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110275138A (en) * | 2019-07-16 | 2019-09-24 | 北京工业大学 | A kind of more sound localization methods removed using advantage sound source ingredient |
| CN115273795A (en) * | 2022-06-22 | 2022-11-01 | 腾讯科技（深圳）有限公司 | Method and device for generating analog impulse response and computer equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105403860A (en) * | 2014-08-19 | 2016-03-16 | 中国科学院声学研究所 | Multi-sparse-sound-source positioning method based on predomination correlation |
| US20160139245A1 (en) * | 2014-05-22 | 2016-05-19 | The United States Of America As Represented By The Secretary Of The Navy | Sparsity-Driven Passive Tracking of Acoustic Sources |
| CN105652233A (en) * | 2015-12-29 | 2016-06-08 | 青岛理工大学 | Random sparse acoustic array super-directional structure generation method |
| CN107247251A (en) * | 2017-06-20 | 2017-10-13 | 西北工业大学 | Three-dimensional sound localization method based on compressed sensing |
-
2018
- 2018-06-25 CN CN201810660554.2A patent/CN109001680A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160139245A1 (en) * | 2014-05-22 | 2016-05-19 | The United States Of America As Represented By The Secretary Of The Navy | Sparsity-Driven Passive Tracking of Acoustic Sources |
| CN105403860A (en) * | 2014-08-19 | 2016-03-16 | 中国科学院声学研究所 | Multi-sparse-sound-source positioning method based on predomination correlation |
| CN105652233A (en) * | 2015-12-29 | 2016-06-08 | 青岛理工大学 | Random sparse acoustic array super-directional structure generation method |
| CN107247251A (en) * | 2017-06-20 | 2017-10-13 | 西北工业大学 | Three-dimensional sound localization method based on compressed sensing |
Non-Patent Citations (3)
| Title |
|---|
| AFSANEH ASAEI: "MODEL-BASED SPARSE COMPONENT ANALYSIS FOR MULTIPARTY DISTANT SPEECH RECOGNITION", 《EPFL》 * |
| ANGSHUL MAJUMDAR: "Greedy algorithms promoting group sparsity", 《MATH WORKS》 * |
| DEANNA NEEDELL等: "UNIFORM UNCERTAINTY PRINCIPLE AND SIGNAL RECOVERY VIA REGULARIZED ORTHOGONAL MATCHING PURSUIT", 《MATHEMATICS》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110275138A (en) * | 2019-07-16 | 2019-09-24 | 北京工业大学 | A kind of more sound localization methods removed using advantage sound source ingredient |
| CN110275138B (en) * | 2019-07-16 | 2021-03-23 | 北京工业大学 | Multi-sound-source positioning method using dominant sound source component removal |
| CN115273795A (en) * | 2022-06-22 | 2022-11-01 | 腾讯科技（深圳）有限公司 | Method and device for generating analog impulse response and computer equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108828525A (en) | The more sonic location systems of distributed microphone array in reverberation room | |
| TWI556654B (en) | Apparatus and method for deriving a directional information and systems | |
| CN108802687A (en) | The more sound localization methods of distributed microphone array in reverberation room | |
| US9711126B2 (en) | Methods, systems, and computer readable media for simulating sound propagation in large scenes using equivalent sources | |
| CN109254266A (en) | Sound localization method, device and storage medium based on microphone array | |
| CN107167770B (en) | A kind of microphone array sound source locating device under the conditions of reverberation | |
| Jin et al. | Multizone soundfield reproduction using orthogonal basis expansion | |
| CN109001681A (en) | The method of construction compression observing matrix in more auditory localizations | |
| EP4018685A1 (en) | Bidirectional propagation of sound | |
| US20210266693A1 (en) | Bidirectional Propagation of Sound | |
| Salvati et al. | Sound source and microphone localization from acoustic impulse responses | |
| CN109001680A (en) | The sparse optimization algorithm of block in auditory localization | |
| CN108828526A (en) | The method that expression of the signal S in space-frequency domain in more auditory localizations in two-dimensional grid has block sparsity | |
| Morgenstern et al. | Joint spherical beam forming for directional analysis of reflections in rooms | |
| Hosseini et al. | Time difference of arrival estimation of sound source using cross correlation and modified maximum likelihood weighting function | |
| Georgiou et al. | Incorporating directivity in the Fourier pseudospectral time-domain method using spherical harmonics | |
| Zhong et al. | Spherical convolutional recurrent neural network for real-time sound source tracking | |
| CN109031203A (en) | The algorithm of the recovery block-sparse signal of the more auditory localizations of distributed microphone array in reverberation room | |
| Chen et al. | Sound Field Estimation around a Rigid Sphere with Physics-informed Neural Network | |
| Pertilä | Acoustic source localization in a room environment and at moderate distances | |
| Dang et al. | Multiple sound source localization based on a multi-dimensional assignment model | |
| Ðurković et al. | Low latency localization of multiple sound sources in reverberant environments | |
| Canclini et al. | Distributed 3D source localization from 2D DOA measurements using multiple linear arrays | |
| Wang et al. | Robust Time-Delay Estimation for Speaker Localization Using Mutual Information Among Multiple Microphone Signals | |
| Jian et al. | Acoustic modal analysis of room responses from the perspective of state-space balanced realization with application to field interpolation |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181214 |
|
| RJ01 | Rejection of invention patent application after publication |