CN109283487A - MUSIC-DOA method based on the response of support vector machines controlled power - Google Patents
MUSIC-DOA method based on the response of support vector machines controlled power Download PDFInfo
- Publication number
- CN109283487A CN109283487A CN201811072707.8A CN201811072707A CN109283487A CN 109283487 A CN109283487 A CN 109283487A CN 201811072707 A CN201811072707 A CN 201811072707A CN 109283487 A CN109283487 A CN 109283487A
- Authority
- CN
- China
- Prior art keywords
- srp
- music
- signal
- estimation
- formula
- 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
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
- 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
- G01S3/802—Systems for determining direction or deviation from predetermined direction
-
- 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
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)
- Circuit For Audible Band Transducer (AREA)
Abstract
The present invention discloses a kind of MUSIC-DOA estimation method based on the response of support vector machines controlled power, includes the following steps: step 1, calculates the mean power of microphone array output;Step 2, the Estimation of Spatial Spectrum of SRP-MUSIC method is normalized, obtains SRP-NMUSIC Estimation of Spatial Spectrum;Step 3, support vector machines is introduced, and is estimated with SVM monitor model;Step 4, it is trained by the biasing of subband signal Estimation of Spatial Spectrum and test sample;Step 5, under the premise of known to the sound source angle, SVM is estimated applied to broadband signal DOA, obtains SRP-WMUSIC method DOA estimated result.Such estimation method can effectively solve the problems, such as that MUSIC method for normalizing DOA evaluated error is larger when low signal-to-noise ratio, be a kind of accurate, robust DOA estimation method.
Description
Technical field
The invention belongs to microphone array DOA estimation technique fields, in particular to a kind of to be based on the controllable function of support vector machines
The MUSIC-DOA method of rate response.
Background technique
Wave beam forming is a kind of steady sound localization method, and the purpose is to by maximizing the space filter on Sounnd source direction
The position of wave device response estimation sound source.Earliest array signal DOA estimation method be proposed by Bartlett in nineteen sixty-five it is normal
Beam-forming schemes, also referred to as Bartlett beam-forming schemes are advised, but this method has " Rayleigh limit ", cannot tell positioned at one
Target in beam angle.Therefore, in practical application, when not increasing physical pore size, how to break through Rayleigh limit is array signal DOA
Estimate an important problem of urgent need to resolve.In the sound localization method based on steerable beam, based on controlled power response and
Phse conversion method strong robustness can obtain more accurate DOA estimation, but in low noise in small noise and appropriate reverberation
Than or when strong reverberation, DOA estimates that accuracy is not high, operand is big.Recent studies suggest that MUSIC method for normalizing can be improved
Spatial resolution can preferably identify straight length and reflection, but also enhance noise component(s), cause DOA evaluated error larger.
Support vector machines replaces empirical risk minimization criterion with empirical risk minimization, is solving small sample, non-
It is linearly with the obvious advantage when the Machine Learning Problems such as (Non Linear).2016, the propositions such as Daniele Salvati utilized RBF
Kernel support vectors machine constructs the undistorted response Beam-former of weighted least-square, and the positioning of near field simple sund source can be effectively treated and ask
Topic.
When in order to solve low signal-to-noise ratio, thus the larger problem of MUSIC method for normalizing DOA evaluated error, this case generates.
Summary of the invention
The purpose of the present invention is to provide a kind of estimation side MUSIC-DOA based on the response of support vector machines controlled power
Method can effectively solve the problems, such as that MUSIC method for normalizing DOA evaluated error is larger when low signal-to-noise ratio, be a kind of accurate, robust
DOA estimation method.
In order to achieve the above objectives, solution of the invention is:
A kind of MUSIC-DOA estimation method based on the response of support vector machines controlled power, includes the following steps:
Step 1, the mean power of microphone array output is calculated;
Step 2, to the Estimation of Spatial Spectrum P of SRP-MUSIC methodSRP-MUSIC(f, θ) is normalized, and obtains SRP-
NMUSIC Estimation of Spatial Spectrum;
Step 3, support vector machines is introduced, and is estimated with SVM monitor model;
Step 4, it is trained by the biasing of subband signal Estimation of Spatial Spectrum and test sample;
Step 5, under the premise of known to the sound source angle, SVM is estimated applied to broadband signal DOA, obtains SRP-
WMUSIC method DOA estimated result.
The detailed content of above-mentioned steps 1 is: setting microphone array includes N number of array element, it is assumed that an azimuth is θsIt is narrow
Band signal is incident in N number of array element, and s (t) is t moment source emission signal, the reception signal of n-th of array element are as follows:
xn(t)=δns(t-τn)+vn(t) n=1,2 ..., N
In formula, δnIndicate gain of n-th of the array element to signal, τnIndicate that signal reaches n-th of array element relative to reference array element
Time delay, vn(t) noise of n-th of array element of t moment is indicated;
T becomes the signal vector X that length is L at the time of will be givenn(t), it may be assumed that
Xn(t)=[xn(t),xn(t-1),...,xn(t-L+1)]T
In formula, ()TRepresenting matrix transposition fortune side accords with, then the reception signal of n-th of array element are as follows:
In formula, j indicates imaginary part, frequency domain broadband signal model are as follows:
It is weighted summation by receiving signal to array element, obtains the array output vector that direction of observation is θ, frequency is f
Are as follows:
In formula, H indicates Hermitian transposition, and w (f, θ) is direction vector of the array on the θ of direction;
The mean power of array output are as follows:
P (f, θ)=E | Y (f, θ) |2}=w (f, θ)HΦ(f)w(f,θ)
In formula, Φ (f)=E { x (k, f) xH(k, f) } be array output signal covariance matrix.
The particular content of above-mentioned steps 2 is: SRP-MUSIC Estimation of Spatial Spectrum are as follows:
In formula, G (f) indicates that noise subspace M × (M-N) eigenvectors matrix, a (f, θ) indicate the battle array of even linear array
Column direction vector;
To PSRP-MUSIC(f, θ) is normalized, and obtains SRP-NMUSIC Estimation of Spatial Spectrum, it may be assumed that
In formula, max [] indicates maximum value.
In above-mentioned steps 2, the array direction vector of even linear array in far field are as follows:
In formula, c indicates the velocity of sound, and d indicates array element spacing;
According to MUSIC principle, under ideal conditions, signal subspace and noise subspace in data space are mutually just
It hands over, it may be assumed that
a(f,θ)HUN=0
In formula, USIt is the subspace opened by the corresponding feature vector of N number of characteristic value of covariance matrix Φ, that is, believes
Work song space;UNIt is the subspace opened by the corresponding feature vector of M-N small characteristic value of covariance matrix Φ, i.e. noise
Space.
In above-mentioned steps 3, SRP-WMUSIC broadband signal Estimation of Spatial Spectrum are as follows:
In formula, PSRP-MUSIC(f, θ) indicates SRP-MUSIC Estimation of Spatial Spectrum, fminAnd fmaxIndicate the frequency model of broadband signal
It encloses, γfIt is binary variable, value is -1 or 1, is estimated with SVM monitor model are as follows:
In formula, sgn indicates that jump function, Q indicate training set size, and f is the frequency of subband,It is
I-th of training set sample that frequency is f inputsWithGaussian kernel function,It is the target value of i-th of training set sample,
γfValue be { -1,1 }, ai>=0, b are constants.
In above-mentioned steps 4, biasing is defined as:
In formula, μ (f) is all subband PSRP-NMUSICThe average of (f, θ).
The particular content of above-mentioned steps 5 is: it is θ that the Unite States Standard noise of reference, which is placed on angle,tPosition on, SRP-
NMUSIC method DOA estimation are as follows:
In formula, PSRP-NMUSIC(f, θ) indicates SRP-WMUSIC broadband signal Estimation of Spatial Spectrum;
Sound source angle and SRP-NMUSIC method DOA evaluated error are as follows:
The sample label of i-th of training setIs defined as:
In formula, η is a given threshold value, for distinguishing the subband signal of MUSIC normalization Estimation of Spatial Spectrum inaccuracy,
The subband signal label of inaccuracy is -1, and accurate subband signal label is 1;
Finally, SRP-WMUSIC method DOA estimates are as follows:
In formula, PSRP-WMUSICThe Estimation of Spatial Spectrum of (f, θ) expression SRP-WMUSIC method.
After adopting the above scheme, the present invention carries out Fast Fourier Transform (FFT) to broadband signal first, then uses MUSIC method
DOA estimation is carried out to each subband signal, is classified finally by DOA estimated result of the SVM to subband signal, selection sort
The accurate subband signal of DOA estimated result is merged afterwards, obtains the DOA estimation of broadband signal.The present invention is supervised in audio
Control, radar, sonar, Remote Video Conference environment and robot sense of hearing etc. have many applications.
Detailed description of the invention
Fig. 1 is the information source spatial spectrum that azimuth is -40 °;
Fig. 2 is dead room environment panorama;
Fig. 3 is anechoic room experimental layout;
Fig. 4 is the root-mean-square error of four kinds of methods with signal-to-noise ratio variation diagram;
Fig. 5 is the root-mean-square error of four kinds of methods with FFT hits variation diagram;
Fig. 6 is the root-mean-square error of four kinds of methods with array number variation diagram;
Fig. 7 is the principle of the present invention figure.
Specific embodiment
Below with reference to attached drawing, technical solution of the present invention and beneficial effect are described in detail.
As shown in fig. 7, the present invention provides a kind of estimation side MUSIC-DOA based on the response of support vector machines controlled power
Method is classified according to DOA estimated result of the SVM Nonlinear Mapping to each subband signal, and DOA estimated result is accurate
Subband signal retains and merges to obtain the DOA estimation of broadband signal, is improved the noise robustness of positioning system.
Assuming that an azimuth is θsNarrow band signal be incident in N number of array element, s (t) be t moment source emission signal,
The reception signal of n-th of array element are as follows:
xn(t)=δns(t-τn)+vn(t) n=1,2 ..., N (1)
In formula, δnIndicate gain of n-th of the array element to signal, τnIndicate that signal reaches n-th of array element relative to reference array element
Time delay, vn(t) noise of n-th of array element of t moment is indicated.
T becomes the signal vector X that length is L at the time of will be givenn(t), it may be assumed that
Xn(t)=[xn(t),xn(t-1),...,xn(t-L+1)]T (2)
In formula, ()TRepresenting matrix transposition fortune side accords with, then the reception signal of n-th of array element are as follows:
In formula, j indicates imaginary part, frequency domain broadband signal model are as follows:
It is weighted summation by receiving signal to array element, obtains the array output vector that direction of observation is θ, frequency is f
Are as follows:
In formula, H indicates Hermitian (complex conjugate) transposition, and w (f, θ) is direction vector of the array on the θ of direction.
The mean power of array output are as follows:
P (f, θ)=E | Y (f, θ) |2}=w (f, θ)HΦ(f)w(f,θ) (6)
In formula, Φ (f)=E { x (k, f) xH(k, f) } be array output signal covariance matrix.Select first array element
As reference, it is assumed that all array elements are all omnidirectionals, identical.Then in far field even linear array array direction vector are as follows:
In formula, c indicates the velocity of sound, and d indicates array element spacing.
MUSIC method is a kind of by carrying out feature decomposition to signal covariance matrix, obtains progressive unbiased orientation and estimates
Meter.MUSIC method has higher accuracy than conventional beamformer method and Capon Minimum Variance method.
Since signal source and noise are independent from each other, signal covariance matrix can be decomposed into signal and noise two
Point, it may be assumed that
In formula, USIt is the subspace opened by the corresponding feature vector of N number of characteristic value of covariance matrix Φ, that is, believes
Work song space;UNIt is the subspace opened by the corresponding feature vector of M-N small characteristic value of covariance matrix Φ, i.e. noise
Space.
According to MUSIC principle, under ideal conditions, signal subspace and noise subspace in data space are mutually just
It hands over, that is, the guiding vector and noise subspace of signal are mutually orthogonal, it may be assumed that
a(f,θ)HUN=0
(9)
It follows that MUSIC method (SRP-MUSIC) Estimation of Spatial Spectrum based on controlled power response are as follows:
In formula, G (f) indicates noise subspace M × (M-N) eigenvectors matrix.
To the Estimation of Spatial Spectrum P of SRP-MUSIC methodSRP-MUSIC(f, θ) is normalized, and obtains ringing based on controlled power
MUSIC normalization (SRP-NMUSIC) Estimation of Spatial Spectrum answered, it may be assumed that
In formula, max [] indicates maximum value.
MUSIC-DOA estimation method based on the response of support vector machines controlled power
When in order to solve low signal-to-noise ratio, MUSIC normalize DOA estimation method degradation problem, here introduce support to
After amount machine, MUSIC (SRP-WMUSIC) the broadband signal Estimation of Spatial Spectrum responded based on support vector machines controlled power is obtained are as follows:
In formula, fminAnd fmaxIndicate the frequency range of broadband signal, γfIt is binary variable, value is -1 or 1, uses SVM
Monitor model is estimated are as follows:
In formula, sgn indicates that jump function, Q indicate training set size, and f is the frequency of subband,It is
I-th of training set sample that frequency is f inputsWithGaussian kernel function,It is the target value of i-th of training set sample,
γfValue be { -1,1 }, ai>=0, b are constants.Parameter a is obtained by solving following convex optimization problemi。
In formula, λ is Study first, influences the distance between supporting vector and separate confinement.With sequence minimum optimization method
It solves equation (14), obtains parameter b are as follows:
Trained and test sample is obtained by the biasing of subband signal Estimation of Spatial Spectrum.Biasing is the degree of being distributed symmetrically property
Amount, is defined as:
In formula, μ (f) is all subband PSRP-NMUSICThe average of (f, θ).
SVM is estimated applied to broadband signal DOA, under the premise of need to establishing known to the sound source angle, by the U.S. of reference
Standard noise (USASI), being placed on angle is θtPosition on.SRP-NMUSIC method DOA estimation are as follows:
Sound source angle and SRP-NMUSIC method DOA evaluated error are as follows:
The sample label of i-th of training setIs defined as:
In formula, η is a given threshold value, for distinguishing the subband signal of MUSIC normalization Estimation of Spatial Spectrum inaccuracy,
The subband signal label of inaccuracy is -1, and accurate subband signal label is 1.
Finally, SRP-WMUSIC method DOA estimates are as follows:
In formula, PSRP-WMUSICThe Estimation of Spatial Spectrum of (f, θ) expression SRP-WMUSIC method.
Effect of the invention is illustrated by following emulation and experiment:
Simulated conditions are as follows: array number M=4, and array element spacing is 5cm, velocity of sound c=340m/s.Training stage uses sampling
The USASI noise signal that frequency is 44.1kHz is as incoming signal, and incident angle is -40 °, signal-to-noise ratio 0dB.Threshold value η setting
It is 3, available preferable SVM training pattern.Test phase uses the human voice signal that sample frequency is 1.6kHz as incident
Signal, signal-to-noise ratio are -5dB, keep identical array structure in trained and test phase, the gaussian kernel function in SVM is according to friendship
Fork proof method sets σ=1 and λ=1.Simulation result is as shown in Figure 1.
As shown in Figure 1, the root-mean-square error of SRP-WMUSIC is 0.7, SRP-NMUSIC when signal-to-noise ratio is -5dB,
The root-mean-square error of SRP-MUSIC and SRP-PHAT is respectively 1.7,2 and 4.This paper SRP-WMUSIC method is excellent in accuracy
In SRP-NMUSIC, the methods of SRP-MUSIC and SRP-PHAT.
Experiment condition and result are as figures 2-6.
As shown in Fig. 2, installing a linear microphone array array structure, in anechoic room for picking up the space of space voice
Information;Placed the speaker in pickup system equipment, provide sound source for microphone array, noise elimination chamber size be 5.5m × 3.3m ×
2.3m。
As shown in figure 3, the array number of microphone array is 4, array element spacing is 5cm.With first array element of microphone array
For reference array element, sound source distance microphone 1.5m.A height of 1.2m of sound source and microphone.Incoming signal angle is respectively 0 °,
10 °, 30 °, 45 ° and 60 °.
As shown in figure 4, solid line indicates experimental result, dotted line indicates simulation result.It is testing in compared with emulating, with
The increase of signal-to-noise ratio, SRP-WMUSIC, SRP-NMUSIC, the methods of SRP-MUSIC and SRP-PHAT experimental result and emulation knot
The root-mean-square error of fruit is all gradually reduced, but since there are electrical noises in microphone, so that the root-mean-square error of experimental result is equal
Higher than simulation result.In experimental result, with the increase of signal-to-noise ratio, the root-mean-square error of this paper SRP-WMUSIC method is less than
The methods of SRP-NMUSIC, SRP-MUSIC and SRP-PHAT have better accuracy and stability and consistent with simulation result.
As shown in figure 5, testing in compared with emulating, with the increase of FFT hits, SRP-WMUSIC, SRP-
NMUSIC and SRP-MUSIC methods experiment result and the root-mean-square error of simulation result are all gradually reduced, and the root mean square of experiment misses
Difference is above simulation result, and mainly as caused by electrical noise in microphone, the root mean square of SRP-PHAT methods experiment result is missed
Difference gradually increases after hits reaches 256, becomes larger with simulation result gap, is mainly missed by array element phase in experiment
Difference is affected to SRP-PHAT method.In experimental result, when FFT hits is smaller, cause the sample number of test set compared with
It is small, make SRP-PHAT method accuracy better than SRP-WMUSIC method, with the increase of FFT hits, this paper SRP-WMUSIC
The root-mean-square error of method is gradually reduced and accuracy is better than SRP-NMUSIC, the methods of SRP-MUSIC and SRP-PHAT, and imitative
True result is consistent.
As shown in fig. 6, testing in compared with emulating, SRP-WMUSIC, SRP-NMUSIC, SRP-MUSIC and SRP-
The root-mean-square error of the methods of PHAT experimental result is above simulation result, but with the increase of element number of array, signal covariance
The spatial signal information amount that matrix includes is richer, so that the root-mean-square error of experiment and emulation all gradually decreases.It is tied in experiment
In fruit, with the increase of element number of array, the root-mean-square error of this paper SRP-WMUSIC method is gradually reduced and better than SRP-
The methods of NMUSIC, SRP-MUSIC and SRP-PHAT, it is consistent with simulation result.
The above examples only illustrate the technical idea of the present invention, and this does not limit the scope of protection of the present invention, all
According to the technical idea provided by the invention, any changes made on the basis of the technical scheme each falls within the scope of the present invention
Within.
Claims (7)
1. a kind of MUSIC-DOA estimation method based on the response of support vector machines controlled power, it is characterised in that including walking as follows
It is rapid:
Step 1, the mean power of microphone array output is calculated;
Step 2, to the Estimation of Spatial Spectrum P of SRP-MUSIC methodSRP-MUSIC(f, θ) is normalized, and obtains SRP-NMUSIC sky
Between Power estimation;
Step 3, support vector machines is introduced, and is estimated with SVM monitor model;
Step 4, it is trained by the biasing of subband signal Estimation of Spatial Spectrum and test sample;
Step 5, under the premise of known to the sound source angle, SVM is estimated applied to broadband signal DOA, obtains the side SRP-WMUSIC
Method DOA estimated result.
2. the MUSIC-DOA estimation method as described in claim 1 based on the response of support vector machines controlled power, feature exist
In: the detailed content of the step 1 is: setting microphone array includes N number of array element, it is assumed that an azimuth is θsNarrow band signal
It is incident in N number of array element, s (t) is t moment source emission signal, the reception signal of n-th of array element are as follows:
xn(t)=δns(t-τn)+vn(t) n=1,2 ..., N
In formula, δnIndicate gain of n-th of the array element to signal, τnIndicate signal reach n-th of array element relative to reference array element when
Prolong, vn(t) noise of n-th of array element of t moment is indicated;
T becomes the signal vector X that length is L at the time of will be givenn(t), it may be assumed that
Xn(t)=[xn(t),xn(t-1),...,xn(t-L+1)]T
In formula, ()TRepresenting matrix transposition fortune side accords with, then the reception signal of n-th of array element are as follows:
In formula, j indicates imaginary part, frequency domain broadband signal model are as follows:
It is weighted summation by receiving signal to array element, obtains the array output vector that direction of observation is θ, frequency is f are as follows:
In formula, H indicates Hermitian transposition, and w (f, θ) is direction vector of the array on the θ of direction;
The mean power of array output are as follows:
P (f, θ)=E | Y (f, θ) |2}=w (f, θ)HΦ(f)w(f,θ)
In formula, Φ (f)=E { x (k, f) xH(k, f) } be array output signal covariance matrix.
3. the MUSIC-DOA estimation method as described in claim 1 based on the response of support vector machines controlled power, feature exist
In: the particular content of the step 2 is: SRP-MUSIC Estimation of Spatial Spectrum are as follows:
In formula, G (f) indicates that noise subspace M × (M-N) eigenvectors matrix, a (f, θ) indicate the array side of even linear array
To vector;
To PSRP-MUSIC(f, θ) is normalized, and obtains SRP-NMUSIC Estimation of Spatial Spectrum, it may be assumed that
In formula, max [] indicates maximum value.
4. the MUSIC-DOA estimation method as claimed in claim 3 based on the response of support vector machines controlled power, feature exist
In: in the step 2, the array direction vector of even linear array in far field are as follows:
In formula, c indicates the velocity of sound, and d indicates array element spacing;
According to MUSIC principle, under ideal conditions, signal subspace in data space is mutually orthogonal with noise subspace
, it may be assumed that
a(f,θ)HUN=0
In formula, USIt is the subspace opened by the corresponding feature vector of N number of characteristic value of covariance matrix Φ, that is, signal subspace
Space;UNIt is the subspace opened by the corresponding feature vector of M-N small characteristic value of covariance matrix Φ, i.e. noise is empty
Between.
5. the MUSIC-DOA estimation method as described in claim 1 based on the response of support vector machines controlled power, feature exist
In: in the step 3, SRP-WMUSIC broadband signal Estimation of Spatial Spectrum are as follows:
In formula, PSRP-MUSIC(f, θ) indicates SRP-MUSIC Estimation of Spatial Spectrum, fminAnd fmaxIndicate the frequency range of broadband signal,
γfIt is binary variable, value is -1 or 1, is estimated with SVM monitor model are as follows:
In formula, sgn indicates that jump function, Q indicate training set size, and f is the frequency of subband,It is frequency
It is inputted for i-th of training set sample of fWithGaussian kernel function,It is the target value of i-th of training set sample, γf's
Value is { -1,1 }, ai>=0, b are constants.
6. the MUSIC-DOA estimation method as described in claim 1 based on the response of support vector machines controlled power, feature exist
In: in the step 4, biasing is defined as:
In formula, μ (f) is all subband PSRP-NMUSICThe average of (f, θ).
7. the MUSIC-DOA estimation method as described in claim 1 based on the response of support vector machines controlled power, feature exist
In: the particular content of the step 5 is: it is θ that the Unite States Standard noise of reference, which is placed on angle,tPosition on, SRP-NMUSIC
Method DOA estimation are as follows:
In formula, PSRP-NMUSIC(f, θ) indicates SRP-WMUSIC broadband signal Estimation of Spatial Spectrum;
Sound source angle and SRP-NMUSIC method DOA evaluated error are as follows:
The sample label of i-th of training setIs defined as:
In formula, η is a given threshold value, for distinguishing the subband signal of MUSIC normalization Estimation of Spatial Spectrum inaccuracy, is not allowed
True subband signal label is -1, and accurate subband signal label is 1;
Finally, SRP-WMUSIC method DOA estimates are as follows:
In formula, PSRP-WMUSICThe Estimation of Spatial Spectrum of (f, θ) expression SRP-WMUSIC method.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810437771 | 2018-05-09 | ||
CN2018104377715 | 2018-05-09 |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109283487A true CN109283487A (en) | 2019-01-29 |
CN109283487B CN109283487B (en) | 2023-04-21 |
Family
ID=65181271
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811072707.8A Active CN109283487B (en) | 2018-05-09 | 2018-09-14 | MUSIC-DOA method based on controllable power response of support vector machine |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109283487B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112965026A (en) * | 2021-02-08 | 2021-06-15 | 华南理工大学 | DOA array element spacing setting method with prior positioning angle range |
CN114639398A (en) * | 2022-03-10 | 2022-06-17 | 电子科技大学 | Broadband DOA estimation method based on microphone array |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103091661A (en) * | 2013-02-01 | 2013-05-08 | 西安科技大学 | Broadband signal arriving direction estimation method based on iteration spectral reconfiguration |
CN104698433A (en) * | 2015-03-16 | 2015-06-10 | 电子科技大学 | Single-snapshot data-based coherent signal DOA (direction of arrival) estimating method |
CN106443570A (en) * | 2016-08-22 | 2017-02-22 | 西安电子科技大学 | Direction of arrival estimation method based on multiple signal classification algorithm vector correlation |
CN107907852A (en) * | 2017-10-27 | 2018-04-13 | 大连大学 | Covariance matrix order based on space smoothing minimizes DOA estimation method |
KR102183439B1 (en) * | 2019-05-29 | 2020-11-26 | 부산대학교 산학협력단 | Method and apparatus for estimating direction of arrival using combined beamspace music and tma |
-
2018
- 2018-09-14 CN CN201811072707.8A patent/CN109283487B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103091661A (en) * | 2013-02-01 | 2013-05-08 | 西安科技大学 | Broadband signal arriving direction estimation method based on iteration spectral reconfiguration |
CN104698433A (en) * | 2015-03-16 | 2015-06-10 | 电子科技大学 | Single-snapshot data-based coherent signal DOA (direction of arrival) estimating method |
CN106443570A (en) * | 2016-08-22 | 2017-02-22 | 西安电子科技大学 | Direction of arrival estimation method based on multiple signal classification algorithm vector correlation |
CN107907852A (en) * | 2017-10-27 | 2018-04-13 | 大连大学 | Covariance matrix order based on space smoothing minimizes DOA estimation method |
KR102183439B1 (en) * | 2019-05-29 | 2020-11-26 | 부산대학교 산학협력단 | Method and apparatus for estimating direction of arrival using combined beamspace music and tma |
Non-Patent Citations (3)
Title |
---|
李新等: "基于Root-MUSIC和支持向量机的间谐波参数估计", 《电测与仪表》 * |
杜金香等: "支持向量机在DOA估计中的稳健性研究", 《声学技术》 * |
龚珞珞: "基于压缩感知的MIMO雷达角度估计方法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112965026A (en) * | 2021-02-08 | 2021-06-15 | 华南理工大学 | DOA array element spacing setting method with prior positioning angle range |
CN112965026B (en) * | 2021-02-08 | 2023-07-18 | 华南理工大学 | DOA array element spacing setting method with priori positioning angle range |
CN114639398A (en) * | 2022-03-10 | 2022-06-17 | 电子科技大学 | Broadband DOA estimation method based on microphone array |
CN114639398B (en) * | 2022-03-10 | 2023-05-26 | 电子科技大学 | Broadband DOA estimation method based on microphone array |
Also Published As
Publication number | Publication date |
---|---|
CN109283487B (en) | 2023-04-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109272989B (en) | Voice wake-up method, apparatus and computer readable storage medium | |
Mohan et al. | Localization of multiple acoustic sources with small arrays using a coherence test | |
CN109102822B (en) | Filtering method and device based on fixed beam forming | |
CN107976651B (en) | Sound source positioning method and device based on microphone array | |
CN105301563B (en) | A kind of double sound source localization method that least square method is converted based on consistent focusing | |
Salvati et al. | Incoherent frequency fusion for broadband steered response power algorithms in noisy environments | |
Sun et al. | Joint DOA and TDOA estimation for 3D localization of reflective surfaces using eigenbeam MVDR and spherical microphone arrays | |
Grondin et al. | Time difference of arrival estimation based on binary frequency mask for sound source localization on mobile robots | |
CN111044973A (en) | MVDR target sound source directional pickup method for microphone matrix | |
Salvati et al. | On the use of machine learning in microphone array beamforming for far-field sound source localization | |
CN108549052A (en) | A kind of humorous domain puppet sound intensity sound localization method of circle of time-frequency-spatial domain joint weighting | |
CN106371057B (en) | Voice sound source direction-finding method and device | |
CN106019230B (en) | A kind of sound localization method based on i-vector Speaker Identification | |
Pujol et al. | BeamLearning: An end-to-end deep learning approach for the angular localization of sound sources using raw multichannel acoustic pressure data | |
CN109283487A (en) | MUSIC-DOA method based on the response of support vector machines controlled power | |
Yen et al. | Multi-sensory sound source enhancement for unmanned aerial vehicle recordings | |
Meier et al. | Artificial Neural Network-Based Feature Combination for Spatial Voice Activity Detection. | |
CN110047507A (en) | A kind of identification of sound source method and device | |
CN116559778B (en) | Vehicle whistle positioning method and system based on deep learning | |
Thomas et al. | Optimal 3D beamforming using measured microphone directivity patterns | |
Yamaoka et al. | Performance evaluation of nonlinear speech enhancement based on virtual increase of channels in reverberant environments | |
Hadad et al. | Multi-speaker direction of arrival estimation using SRP-PHAT algorithm with a weighted histogram | |
Moore et al. | 2D direction of arrival estimation of multiple moving sources using a spherical microphone array | |
Beit-On et al. | Binaural direction-of-arrival estimation in reverberant environments using the direct-path dominance test | |
Chen et al. | A DNN based normalized time-frequency weighted criterion for robust wideband DoA estimation |
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 | ||
CB02 | Change of applicant information |
Address after: 210032 No. 219 Ning six road, Jiangbei new district, Nanjing, Jiangsu Applicant after: Nanjing University of Information Science and Technology Address before: 211500 Yuting Square, 59 Wangqiao Road, Nanjing, Jiangsu Province Applicant before: Nanjing University of Information Science and Technology |
|
CB02 | Change of applicant information | ||
GR01 | Patent grant | ||
GR01 | Patent grant |