CN109584900A - A kind of blind source separation algorithm of signals and associated noises - Google Patents
A kind of blind source separation algorithm of signals and associated noises Download PDFInfo
- Publication number
- CN109584900A CN109584900A CN201811357511.3A CN201811357511A CN109584900A CN 109584900 A CN109584900 A CN 109584900A CN 201811357511 A CN201811357511 A CN 201811357511A CN 109584900 A CN109584900 A CN 109584900A
- Authority
- CN
- China
- Prior art keywords
- signal
- observation signal
- original observation
- original
- 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.)
- Pending
Links
- 238000000926 separation method Methods 0.000 title claims abstract description 37
- 239000011159 matrix material Substances 0.000 claims abstract description 32
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 22
- 238000000034 method Methods 0.000 claims abstract description 18
- 238000001914 filtration Methods 0.000 claims abstract description 7
- 230000002087 whitening effect Effects 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 5
- 238000005516 engineering process Methods 0.000 abstract description 2
- 230000007547 defect Effects 0.000 description 2
- 238000012880 independent component analysis Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 239000000470 constituent Substances 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000005236 sound signal Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Processing of the speech or voice signal to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0272—Voice signal separating
- G10L21/028—Voice signal separating using properties of sound source
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Processing of the speech or voice signal to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Processing of the speech or voice signal to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
- G10L21/0216—Noise filtering characterised by the method used for estimating noise
- G10L21/0232—Processing in the frequency domain
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Processing of the speech or voice signal to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0272—Voice signal separating
- G10L21/0308—Voice signal separating characterised by the type of parameter measurement, e.g. correlation techniques, zero crossing techniques or predictive techniques
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
Abstract
The invention belongs to signal processing technology fields, are related to a kind of blind source separation algorithm of signals and associated noises, comprising: step 1, input original observation signal, pre-process to the original observation signal；Step 2: pretreated original observation signal being filtered using wavelet packet, threshold value is then selected and the original observation signal after decomposition is reconstructed to obtain filtered observation signal using wavelet packet；Step 3: establishing cost function using maximum signal noise ratio principle；Step 4: generalized eigenvalue decomposition being carried out to cost function and obtains separation matrix, then obtains each source signal with original observation signal matrix multiple；Step 5: half sliding filtering being carried out to each source signal described in step 4, the source signal after finally obtaining noise reduction.The method of the present invention calculates simply for having preferable denoising effect compared with the signal under low signal-to-noise ratio, and the speed of service is fast, has biggish use value in terms of Speech processing.
Description
Technical field
The invention belongs to signal processing technology fields, are related to a kind of blind source separation algorithm of signals and associated noises.
Background technique
Blind source separating is the emerging field to grow up last century end, is artificial neural network, statistic line loss rate
And the product that information theory combines, great application has been obtained in many fields.Blind source separating is exactly not know
In the case where the constituent and transmission channel of source signal, only restore or isolate the mistake of each source signal from observation signal
Journey needs to look for a separation matrix, each source signal is extracted from observation signal.
Occurring in this more than 20 years so far in blind source separating, domestic and international researcher proposes various separation algorithms,
And it cuts both ways.
Such as based on the blind source separation algorithm of independent component analysis, it is the very classical effective side in blind source separation algorithm
Method, it needs supposed premise condition, i.e., source signal meets non-gaussian distribution and mutually indepedent.Secondly there is also some disadvantages for the algorithm
End, first is that the signal after separation can not be allowed to correspond with the signal before mixing；Second is that can not restore to mix the true of front signal
Amplitude.These disadvantages be for the purpose of final separation within an acceptable range, it is right but in certain practical applications
Semaphore request after separation is harsh, so these defects are also that can not ignore.
There are also some algorithms emerging in recent years, and such as sparse component analysis, it can fail in independent component analysis method
In the case where, i.e., source signal be unsatisfactory for Gaussian Profile or it is mutually indepedent in the case where, observation signal is separated.For another example non-
Negative matrix decomposition algorithm, it is mutually indepedent also without information source and meets the limitation of non-gaussian distribution, only increases decomposition square
The non-negative constraint of element in battle array.These above-mentioned methods are mostly to establish cost function with certain criterion, are then iterated to it
Optimize and obtain separation matrix, but requires that signal has enough degree of rarefications.
Stone J V proposes the blind source separation algorithm based on time domain prediction, and the ratio of two errors of the algorithm constructs
Objective function, the i.e. long anticipation component of signal make molecule, and the short anticipation component of signal makees denominator, then carry out to objective function excellent
Change, i.e., generalized eigenvalue solves.Because this kind of algorithm has lower computation complexity, Cheung Y M et al. is in this base later
The blind source separation algorithm of global optimization is had also been proposed on plinth, Zhang little Bing et al. has also been proposed based on maximum noise on this basis
The blind source separation algorithm of ratio.But the algorithm have the defects that it is certain: first is that the bigger separating effect of signal-to-noise ratio is better, in lower letter
It makes an uproar than under, separating effect is deteriorated；Second is that replacing source signal to exist with the sliding average of estimation signal the estimation problem of source signal
It will lead to error to a certain extent or algorithm fail under certain conditions.
Summary of the invention
In view of the deficiencies of the prior art, the present invention discloses one using the thought of double-smoothing from the angle of noise reduction
The blind source separation algorithm of kind signals and associated noises.
To achieve the above object, present invention employs following design structure and design schemes: a kind of signals and associated noises
Blind source separation algorithm comprises the following specific steps that: step 1, inputting original observation signal, is located in advance to the original observation signal
Reason, including go mean value and albefaction；Step 2: being filtered, wrapped using wavelet packet original observation signal pretreated to step 1
It includes and the pretreated original observation signal is decomposed, then select threshold value and using wavelet packet to the original sight after decomposition
Signal is surveyed to be reconstructed to obtain filtered observation signal；Step 3: cost function is established using maximum signal noise ratio principle, wherein
Estimate that filtered observation signal described in the sliding average step 2 of signal replaces；Step 4: broad sense is carried out to cost function
Eigenvalues Decomposition obtains separation matrix, then obtains each source signal with original observation signal matrix multiple；Step 5: to step 4
Each source signal carries out half sliding filtering, the source signal after finally obtaining noise reduction.
Further, it is that the mean value of the original observation signal is subtracted from original observation signal that mean value is gone in the step 1
Vector, so that original observation signal becomes zero-mean variable, method is to assume that certain observation vector x is mean value in original observation signal
The random vector being not zero, then x0The mean value of=x-E (x), E (x) expression x.
Further, the method for albefaction assumes that original observation signal x in the step 10Correlation matrix be Rx, then Rx
Eigenvalues Decomposition be Rx=Q Σ2QT, Σ in formula2For diagonal matrix, diagonal element isFor matrix RxSpy
Value indicative, and the column vector of orthogonal matrix Q is the feature vector of normal orthogonal corresponding with characteristic value, whitening matrix takes T=Σ- 1QT, thenAfter the transformation of whitening matrix T, original observation signalIt is irrelevant between each component.
Further, pretreated original observation signal is filtered in the step 2, is included the following steps:
(1) original observation signal is decomposed using wavelet packet, select a small echo and determines the level decomposed, then
Carry out decomposition computation, calculation method are as follows:In formula,
K ∈ Z,Indicate the obscuring component of signal；Indicate the details coefficients of signal；hk-2iIndicate low-pass filter, with
Scaling function is related；gk-2iIndicate that high-pass filter, g are related with wavelet function；
(2) threshold value of WAVELET PACKET DECOMPOSITION coefficient is chosen, threshold function table is as follows:
Or
Wherein, formula (1) is hard threshold function, and formula (2) is soft-threshold function；In formulaIndicate that the small echo of signal becomes
It changes, Th () indicates threshold function table, and the selection of the threshold value λ in two formulas is usually determined by following formula, i.e.,, formula
In, σnFor noise criteria variance, N is the length of signal；
(3) wavelet package reconstruction is carried out to observation signal, obtains filtered observation signal, method is with rearmost small echo
Subject to packet decomposition coefficient and treated coefficient, wavelet package reconstruction is carried out, i.e.,In formula, h indicates analysis filter, and related with scaling function, g is indicated
Reconfigurable filter,Indicate the obscuring component of signal；Indicate the details coefficients of signal.
Further, the original observation signal is subtracted using the filtered observation signal and obtain noise error, make
Cost function is established with maximum signal noise ratio principle.
Further, when being decomposed using wavelet packet to original observation signal, using dB3 small echo to the original observation
Signal carries out 3 layers of decomposition.
Design principle of the invention are as follows: the present invention first carries out original observation signal pre- by the way of Double-noise-reduction
Processing, including mean value and albefaction are gone, convenient for carrying out calculating analysis to signal, secondly, with the observation signal after wavelet filtering come generation
For the sliding average of estimation signal, cost function is established using this criterion of signal-to-noise ratio maximum when optimal separation effect；Then, right
Cost function carries out generalized eigenvalue decomposition, obtains separation matrix, then obtains each source with original observation signal matrix multiple
Signal；Finally, carrying out smothing filtering to each source signal again obtains the final separation signal under more low signal-to-noise ratio.
The present invention generates following beneficial effect compared with prior art.The method of the present invention has the signal compared with low signal-to-noise ratio
There is preferable denoising effect, and have and calculate simple, the fast feature of the speed of service, has in terms of Speech processing larger
Use value.
Detailed description of the invention
Fig. 1 is the flow chart of the blind source separation algorithm of signals and associated noises of the invention.
Fig. 2 is two-way source signal used by the embodiment of the present invention 1.
Fig. 3 is to carry out mixing obtained original observation signal to two-way source signal in Fig. 2 in the present invention.
Fig. 4 be observation signal original in Fig. 3 is added in the present invention make an uproar after obtained observation signal.
Fig. 5 is that obtained separation signal after blind source separating is carried out to observation signal in Fig. 4.
Fig. 6 is to utilize the method for the present invention obtained final source signal after secondary noise reduction smothing filtering.
Specific embodiment
The present invention will be further described with reference to the accompanying drawings and examples.
As shown in Figure 1, the flow chart of the blind source separation algorithm of signals and associated noises of the invention, the present embodiment have chosen such as Fig. 2
Shown in two-way frequency be 22050Hz, duration 1.3s audio signal as source signal s, be denoted as s=[s1(t),s2(t)]T, and
The two-way source signal use hybrid matrix forMixed, as a result as shown in figure 3, then give this two
The noise that road source signal is 25dB plus signal-to-noise ratio, obtains original observation signal X (t), as shown in Figure 4.
Then, it is carried out to the original observation signal X (t) using a kind of blind source separation algorithm of signals and associated noises of the invention
Denoising.
Firstly, being pre-processed to the original observation signal X (t).If original observation signal is X (t)=[x1(t),x2
(t)]T, when removing average value processing to X (t), implementation process is as follows:T=1,2 ..., n, wherein
x0i(t) i-th of zero-mean component is indicated.The original observation signal X (t) described at this time becomes zero-mean signal X0(t)=[x01(t),
x02(t)]T.Then whitening processing is carried out again, implementation process is as follows: the stochastic variable X of zero-mean0Correlation matrix beMathematic expectaion is sought in wherein E [] expression.To RxIt carries out Eigenvalues Decomposition and obtains Rx
=Q Σ2QT, Σ in formula2For diagonal matrix, diagonal element isThat is RxQ=λ Q, (λ I-Rx) Q=0, wherein I
For unit matrix, λ RxCharacteristic value；Find out the feature vector Q that just can be obtained after λ corresponding to it.Take whitening matrix T=Σ- 1QT, then the signal after albefaction be。
Further, signal noise ratio is established according to signal-to-noise ratio (SNR) Criterion.Fig. 2 source signal s is estimated with it error of signal y
E=s-y is as noise signal, the signal noise ratio of foundationWherein,
The definition of estimation signal y is, if source signal is s (n)=[s1(n),…,sN(n)]T, H is the linear hybrid matrix of N × N rank, then
Observation signal is X (n)=[x1(n),…,xN(n)]T, then the mixed model of signal is X (n)=Hs (n)；Want to recover source letter
Number, a separation matrix W is just had to find out, so that y (n)=WX (n)=WHs (n)=Gs (n), the y (n) obtained at this time is exactly s
(n) one estimates that G is global change's matrix in formula.
Since source signal s is unknown, and estimating signal y is to go with noise, therefore with wavelet packet to estimation signal y
The Y obtained after making an uproar replaces source signal s, and above formula becomes at this time
Y in order to obtain, implementation process are as follows: being filtered noise reduction using wavelet packet, include the following steps: that (1) utilizes
Wavelet packet decomposes original observation signal, selects a small echo and determines the level decomposed, then carries out decomposition computation, count
Calculation method are as follows:In formula, k ∈ Z,It indicates
The obscuring component of signal；Indicate the details coefficients of signal；hk-2iIndicate low-pass filter, it is related with scaling function；
gk-2iIndicate that high-pass filter, g are related with wavelet function.
(2) threshold value of WAVELET PACKET DECOMPOSITION coefficient is chosen, threshold function table is as follows:
Or
Wherein, formula (1) is hard threshold function, and formula (2) is soft-threshold function；In formulaIndicate that the small echo of signal becomes
It changes, Th () indicates threshold function table, and the selection of the threshold value λ in two formulas is usually determined by following formula, i.e.,, formula
In, σnFor noise criteria variance, N is the length of signal.
(3) wavelet package reconstruction is carried out to observation signal, obtains filtered observation signal, method is with rearmost small echo
Subject to packet decomposition coefficient and treated coefficient, wavelet package reconstruction is carried out, i.e.,In formula, h indicates analysis filter, and related with scaling function, g is indicated
Reconfigurable filter,Indicate the obscuring component of signal；Indicate the details coefficients of signal.
Further,Y=WX, which is substituted into signal noise ratio, just to be become
In formula；For correlation matrix；V=WCWT、
Further, after obtaining separation matrix W, two source signal components can be respectively obtained by calculating y=WX, as a result as schemed
Shown in 5.
Finally, carrying out adaptive-filtering respectively to two signals again, so that it may the source signal after obtaining Double-noise-reduction, as a result
As shown in Figure 6.
It is different from two signals of Fig. 6 and Fig. 2, is because of the only letter that blind source separation algorithm of the invention restores
Number independence, for source signal amplitude and sequence be it is irreclaimable, in blind source separation algorithm separating resulting amplitude and
The different of sequence is acceptable.Fig. 6 reduces noise for Fig. 5, well believes source for Fig. 3
It number has separated and to have come out.
The above embodiments are merely illustrative of the technical solutions of the present invention, rather than is limited；Although referring to aforementioned reality
Applying example, invention is explained in detail, for those of ordinary skill in the art, still can be to aforementioned implementation
Technical solution documented by example is modified or equivalent replacement of some of the technical features；And these are modified or replace
It changes, the spirit and scope for claimed technical solution of the invention that it does not separate the essence of the corresponding technical solution.
Claims (6)
1. a kind of blind source separation algorithm of signals and associated noises, which is characterized in that comprise the following specific steps that:
Step 1, original observation signal is inputted, which is pre-processed, including go mean value and albefaction；
Step 2: being filtered using wavelet packet original observation signal pretreated to step 1, including pretreated to this
Original observation signal is decomposed, and then selects threshold value and the original observation signal after decomposition is reconstructed using wavelet packet
To filtered observation signal；
Step 3: cost function is established using maximum signal noise ratio principle, wherein described in the sliding average step 2 of estimation signal
Filtered observation signal replaces；
Step 4: to cost function carry out generalized eigenvalue decomposition obtain separation matrix, then with original observation signal matrix multiple
Obtain each source signal；
Step 5: half sliding filtering being carried out to each source signal described in step 4, the source signal after finally obtaining noise reduction.
2. the blind source separation algorithm of signals and associated noises as described in claim 1, which is characterized in that gone in the step 1 mean value be from
The mean vector of the original observation signal is subtracted in original observation signal, so that original observation signal becomes zero-mean variable, side
Method is the random vector that certain observation vector x is mean value is not zero in the original observation signal of hypothesis, then x0=x-E (x), E (x) are indicated
The mean value of x.
3. the blind source separation algorithm of signals and associated noises as described in claim 1, which is characterized in that the method for albefaction in the step 1
Assume that original observation signal x0Correlation matrix be Rx, then RxEigenvalues Decomposition be Rx=Q Σ2QT, Σ in formula2For to angular moment
Battle array, diagonal element areFor matrix RxCharacteristic value, and the column vector of orthogonal matrix Q be it is corresponding with characteristic value
The feature vector of normal orthogonal, whitening matrix take T=Σ-1QT, thenAfter the transformation of whitening matrix T, original observation
SignalIt is irrelevant between each component.
4. the blind source separation algorithm of signals and associated noises as described in claim 1, which is characterized in that in the step 2 to pretreatment after
Original observation signal be filtered, include the following steps:
(1) original observation signal is decomposed using wavelet packet, select a small echo and determines the level decomposed, then carried out
Decomposition computation, calculation method are as follows:In formula, k ∈ Z,Indicate the obscuring component of signal；Indicate the details coefficients of signal；hk-2iLow-pass filter is indicated, with scale letter
Number is related；gk-2iIndicate that high-pass filter, g are related with wavelet function；
(2) threshold value of WAVELET PACKET DECOMPOSITION coefficient is chosen, threshold function table is as follows:
Or
Wherein, formula (1) is hard threshold function, and formula (2) is soft-threshold function；In formulaIndicate the wavelet transformation of signal, Th
() indicates threshold function table, and the selection of the threshold value λ in two formulas is usually determined by following formula, i.e.,, in formula, σn
For noise criteria variance, N is the length of signal；
(3) wavelet package reconstruction is carried out to observation signal, obtains filtered observation signal, method is with rearmost wavelet packet point
It solves subject to coefficient and treated coefficient, carries out wavelet package reconstruction, i.e.,In formula, h indicates analysis filter, and related with scaling function, g is indicated
Reconfigurable filter,Indicate the obscuring component of signal；Indicate the details coefficients of signal.
5. the blind source separation algorithm of signals and associated noises as claimed in claim 4, which is characterized in that believed using the filtered observation
Number subtracting the original observation signal obtains noise error, establishes cost function using maximum signal noise ratio principle.
6. the blind source separation algorithm of signals and associated noises as described in claim 1 or 4, which is characterized in that in utilization wavelet packet to original
When observation signal is decomposed, 3 layers of decomposition are carried out to the original observation signal using dB3 small echo.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811357511.3A CN109584900A (en) | 2018-11-15 | 2018-11-15 | A kind of blind source separation algorithm of signals and associated noises |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811357511.3A CN109584900A (en) | 2018-11-15 | 2018-11-15 | A kind of blind source separation algorithm of signals and associated noises |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109584900A true CN109584900A (en) | 2019-04-05 |
Family
ID=65922478
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811357511.3A Pending CN109584900A (en) | 2018-11-15 | 2018-11-15 | A kind of blind source separation algorithm of signals and associated noises |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109584900A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112364812A (en) * | 2020-11-26 | 2021-02-12 | 上海大学 | aVEPs electroencephalogram identification method based on TRCA-WPTD |
| CN112967722A (en) * | 2021-02-01 | 2021-06-15 | 东北大学 | Text-independent multi-source speaker identification method based on blind source separation |
| CN113823316A (en) * | 2021-09-26 | 2021-12-21 | 南京大学 | Voice signal separation method for sound source close to position |
| CN115001517A (en) * | 2022-05-23 | 2022-09-02 | 四川轻化工大学 | Wireless communication receiving noise suppression method |
| CN116866122A (en) * | 2023-07-13 | 2023-10-10 | 中国人民解放军战略支援部队航天工程大学 | Blind separation method for interference-containing information of transformation domain signal enhancement |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104367315A (en) * | 2014-11-13 | 2015-02-25 | 太原理工大学 | Blind source separation method for extracting J wave signals in electrocardiogram |
| CN104468436A (en) * | 2014-10-13 | 2015-03-25 | 中国人民解放军总参谋部第六十三研究所 | Communication signal wavelet domain blind source separation anti-interference method and device |
-
2018
- 2018-11-15 CN CN201811357511.3A patent/CN109584900A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104468436A (en) * | 2014-10-13 | 2015-03-25 | 中国人民解放军总参谋部第六十三研究所 | Communication signal wavelet domain blind source separation anti-interference method and device |
| CN104367315A (en) * | 2014-11-13 | 2015-02-25 | 太原理工大学 | Blind source separation method for extracting J wave signals in electrocardiogram |
Non-Patent Citations (6)
| Title |
|---|
| JIANCANG MA ET AL.: "Blind Source Separation Algorithm Based on Maximum Signal Noise Ratio", 《2008 FIRST INTERNATIONAL CONFERENCE ON INTELLIGENT NETWORKS AND INTELLIGENT SYSTEMS》 * |
| XIN LIU ET AL.: "Blind Source Separation Algorithm Based on Wavelet Denoising", 《2010 FIRST INTERNATIONAL CONFERENCE ON PERVASIVE COMPUTING, SIGNAL PROCESSING AND APPLICATIONS》 * |
| 张小兵等: "基于最大信噪比的盲源分离算法", 《计算机仿真》 * |
| 王向红等: "基于小波包和独立分量分析的微弱多源鼓掌声发射信号分离", 《上海交通大学学报》 * |
| 申晨宇等: "基于小波包改进的最大信噪比的盲分离算法", 《通信技术》 * |
| 董玮等: "基于最大信噪比的盲源分离雷达抗主瓣干扰方法", 《火力与指挥控制》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112364812A (en) * | 2020-11-26 | 2021-02-12 | 上海大学 | aVEPs electroencephalogram identification method based on TRCA-WPTD |
| CN112364812B (en) * | 2020-11-26 | 2023-04-14 | 上海大学 | aVEPs electroencephalogram identification method based on TRCA-WPTD |
| CN112967722A (en) * | 2021-02-01 | 2021-06-15 | 东北大学 | Text-independent multi-source speaker identification method based on blind source separation |
| CN113823316A (en) * | 2021-09-26 | 2021-12-21 | 南京大学 | Voice signal separation method for sound source close to position |
| CN113823316B (en) * | 2021-09-26 | 2023-09-12 | 南京大学 | Voice signal separation method for sound source close to position |
| CN115001517A (en) * | 2022-05-23 | 2022-09-02 | 四川轻化工大学 | Wireless communication receiving noise suppression method |
| CN116866122A (en) * | 2023-07-13 | 2023-10-10 | 中国人民解放军战略支援部队航天工程大学 | Blind separation method for interference-containing information of transformation domain signal enhancement |
| CN116866122B (en) * | 2023-07-13 | 2024-02-13 | 中国人民解放军战略支援部队航天工程大学 | Blind separation method for interference-containing information of transformation domain signal enhancement |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109584900A (en) | A kind of blind source separation algorithm of signals and associated noises | |
| Zhang et al. | Sparse signal recovery in the presence of correlated multiple measurement vectors | |
| CN101303764B (en) | Method for self-adaption amalgamation of multi-sensor image based on non-lower sampling profile wave | |
| CN111192211B (en) | Multi-noise type blind denoising method based on single deep neural network | |
| CN105913393A (en) | Self-adaptive wavelet threshold image de-noising algorithm and device | |
| CN102944252A (en) | Method for processing fibber Bragg grating (FBG) signals based on translation invariant wavelet | |
| Shemi et al. | Analysis of ECG signal denoising using discrete wavelet transform | |
| CN100417191C (en) | Method of reducing noise for combined images | |
| CN104092447A (en) | Method for constructing biorthogonal wavelet filter bank | |
| CN112084845B (en) | Low-frequency 1/f noise elimination method based on multi-scale wavelet coefficient autocorrelation | |
| CN109709585A (en) | The method for removing coloured noise in GPS coordinate time series | |
| CN116405100A (en) | Distortion signal restoration method based on priori knowledge | |
| CN113066033B (en) | Multi-stage denoising system and method for color image | |
| CN102509268B (en) | Immune-clonal-selection-based nonsubsampled contourlet domain image denoising method | |
| Sugamya et al. | Image enhancement using singular value decomposition | |
| CN103810682A (en) | Novel image fusion method | |
| CN102622731A (en) | Contourlet domain Wiener filtering image denoising method based on two-dimensional Otsu | |
| Dey et al. | Wavelet transform for signal processing in internet-of-Things (IoT) | |
| Ga’fer et al. | Slantlet Transform based Video Denoising | |
| Tang et al. | A method for determining optimal decomposition scale in wavelet threshold de-noising | |
| Nafornita et al. | Estimating directional smoothness of images with the aid of the hyperanalytic wavelet packet transform | |
| uddin Khan et al. | A robust PDE based image de-noising method | |
| Verma et al. | A comparative performance analysis of wavelets in denoising of speech signals | |
| Wang et al. | Study on denoising and enhancement method in SAR image based on wavelet packet and fuzzy set | |
| CN111665542B (en) | Seismic data frequency extension method and system |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190405 |