CN102289715A - Method for adaptively denoising and modeling wavelet neural network based on forward linear prediction (FLP) - Google Patents
Method for adaptively denoising and modeling wavelet neural network based on forward linear prediction (FLP) Download PDFInfo
- Publication number
- CN102289715A CN102289715A CN2011101530251A CN201110153025A CN102289715A CN 102289715 A CN102289715 A CN 102289715A CN 2011101530251 A CN2011101530251 A CN 2011101530251A CN 201110153025 A CN201110153025 A CN 201110153025A CN 102289715 A CN102289715 A CN 102289715A
- Authority
- CN
- China
- Prior art keywords
- signal
- denoising
- flp
- neural network
- wavelet
- 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
Images
Landscapes
- Gyroscopes (AREA)
Abstract
The invention relates to a method for adaptively denoising and modeling a wavelet neural network based on forward linear prediction (FLP). The method comprises the following steps of: (1) performing multi-scale decomposition on a fiber optic gyroscope null-shift signal at a current moment by using wavelet transformation; (2) performing single-branch reconstruction on an approximate signal and a detail signal, which are decomposed, and thus obtaining the reconstructed approximate signal and the reconstructed detail signal; (3) respectively denoising the approximate signal and the detail signal in the step (2) layer by layer by using an FLP method; and (4) training the network by using the signal which is obtained by denoising layer by layer in the step (3) as input of the neural network and using the fiber optic gyroscope signal at the next moment as output, and after training is finished, denoising and modeling the fiber optic gyroscope null-shift signal. By the method, denoising and modeling are integrated, and the modeling accuracy and compensation accuracy of the fiber optic gyroscope null-shift signal are effectively improved. The method is easy to implement.
Description
Technical field
The invention belongs to the signal Processing in the inertial technology field, relate to a kind of zero-drift signal disposal route, be particularly related to a kind of Self-adaptation Wavelet Neural Network denoising modeling method-FLP-WNN algorithm, be applicable to various fibre optic gyroscopes based on forward direction linear prediction (FLP).
Background technology
Optical fibre gyro (FOG) proposed to have obtained great development since the principle scheme from 1976.Optical fibre gyro is to be a kind of novel all solid state gyroscope that base growth is got up with the Sagnac effect, it is a kind of inertial measurement component of mechanical rotating part, but, influence the output characteristics of gyro because its output signal often is subjected to The noise and causes drift.Therefore, how to remove noise, drift is compensated, to improve the drift performance of signal, this is a very important problem in the optical fibre gyro signal Processing direction.
An important indicator weighing fiber optic gyroscope performance is zero stability partially, and also becoming zero drift is drift, and it is normally defined the standard deviation (1 σ) of gyro output angle speed under certain averaging time (as 10s).Generally, drift is determined jointly by drift and noise in the static output of optical fibre gyro.Wherein noise is rendered as a kind of random variation of short-term, and low-frequency fluctuation or tendency that drift typically refers in the output change.In optical fibre gyro, noise is two different notions with drift, and the mechanism that their produce is different with characteristic, also is different to the influence of system performance.Under the situation of not considering the factor of drifting about, the noisiness of optical fibre gyro output contains a statistic processes that is called " random walk ", wherein each output data all by on the statistics independently the time constitute, be incoherent each other.Along with the increase of Measuring Time, the variable of this random walk domination will take place one to true measurement convergent situation gradually, this mean value be called optical fibre gyro zero partially.Therefore in theory, if white noise is only arranged in the optical fibre gyro signal, the inclined to one side stability of the long term zero of optical fibre gyro will diminish along with the increase of averaging time and test duration so, and zero value is also gradually constant partially is zero or is a normal value for it.In fact because environmental perturbation and optical fibre gyro remnants " nonreciprocity " cause drift, the inclined to one side margin of stability of the long term zero that makes optical fibre gyro is fixed in a certain magnitude or a certain scope, increase again and also can not improve zero stability partially averaging time.This because the disturbance that environmental perturbation and optical fibre gyro " nonreciprocity " cause has reflected the long-time variation of gyro output signal, be called " drift ".Therefore the optical fibre gyro signal is handled, just must be eliminated noise and drift, to improve the precision of optical fibre gyro.
The method of traditional elimination bias stability of interference fiber optical gyroscope mostly is proceed step by step, promptly at first eliminates the noise of gyro signal, and then utilizes drift model to compensate.In denoising method, mainly contain wavelet analysis, FLP algorithm, LMS algorithm and various low-pass filters etc.; In the drift modeling method, mainly contain methods such as neural net model establishing, arma modeling, Kalman filtering modeling.Wherein, wavelet transformation, FLP algorithm and neural net model establishing have all obtained good effect.But because the proceed step by step of disposal route makes conversion speed slower, denoising effect is limited, and modeling accuracy is also not high enough.The present invention merges wavelet transformation, FLP algorithm and neural net model establishing, obtained a kind of Self-adaptation Wavelet Neural Network denoising modeling method based on forward direction linear prediction (FLP), obtained the good signal denoising effect, set up high-precision drift model, effectively eliminated the influence of drift the optical fibre gyro signal.
Summary of the invention
Technology of the present invention is dealt with problems: overcome the deficiencies in the prior art, a kind of Self-adaptation Wavelet Neural Network denoising modeling method-FLP-WNN algorithm based on forward direction linear prediction (FLP), this method with the advantages of wavelet transformation, FLP algorithm and neural network algorithm together, can effectively remove noise in the optical fibre gyro signal, improve modeling accuracy, and the program that is easy to realizes.
Technical solution of the present invention: a kind of Self-adaptation Wavelet Neural Network denoising modeling method-FLP-WNN algorithm based on forward direction linear prediction (FLP) mainly comprises following four steps:
(1) utilize wavelet transformation that zero-drift signal is carried out multiple dimensioned decomposition
As wavelet basis zero-drift signal is carried out multiple dimensioned decomposition with the db4 small echo, the decomposition number of plies is n, obtains the coefficient of dissociation of each layer, comprises approximation coefficient and detail coefficients.
(2) to decomposition obtain approach coefficient and detail coefficients is carried out single reconstruct
To obtain after decomposing approach coefficient and detail coefficients is carried out single reconstruct, thereby obtain the approximate signal a after the reconstruct
nWith detail signal d
i(i=1,2, Λ, n).The reconstruct wavelet basis is elected the db4 small echo as.
(3) approximation signal after the reconstruct and detail signal are successively carried out the FLP denoising respectively
Prop up at the list that step (2) obtains on the basis of the approximation signal of reconstruct and detail signal, utilize FLP algorithm pairing approximation signal and detail signal to carry out denoising, obtain after the denoising signal a ' (n) with d '
i(i=1,2, Λ, n).Wherein, the exponent number of FLP algorithm is made as 30, and its step-length selects to follow following formula:
E
j=E[|e
n(n) |].μ wherein
jBe the step-length under the different frequency range, j=1,2, Λ, 6, E
jIt is the average of FLP absolute error in the j frequency range.
(4) network is trained, to obtain the model of zero-drift signal
With the signal a ' after the denoising successively that obtains in the step (3) (n) with d '
i(i=1,2, Λ, n) as the input sample of neural network, below constantly an optical fibre gyro signal as output sample, network is trained, training finishes and promptly obtains the model of zero-drift signal.Wherein output sample is through the gyro signal based on the current time after the adaptive wavelets transform denoising of forward direction linear prediction (FLP)
Principle of the present invention is: when static, the output signal of optical fibre gyro is a random function, considers the long term drift of its mean value, and it is made up of white noise and a function that slowly changes, and is referred to as drift.In the practical application of optical fibre gyro, drift is a key factor that influences the optical fibre gyro precision, therefore must handle the output signal of optical fibre gyro, to eliminate the influence of drift.
Traditional bias stability of interference fiber optical gyroscope disposal route mostly is respectively carries out at noise and drift.At first, can utilize wavelet transformation, FLP algorithm, LMS algorithm and other low-pass filters to carry out denoising at the noise in the optical fibre gyro signal; At eliminating the shifted signal that obtains behind the noise, the general method of setting up mathematical model that adopts is to realizing the compensation to shifted signal, and neural network algorithm, arma modeling and Kalman filtering algorithm etc. have all obtained good, good application.
The present invention combines the processing to noise and drift together, a kind of Self-adaptation Wavelet Neural Network denoising modeling method based on forward direction linear prediction (FLP) has been proposed, at first utilize wavelet transformation that the optical fibre gyro signal is carried out multiple dimensioned decomposition, and each layer coefficients after decomposing carried out individual layer reconstruct, obtain approximation signal a
nWith detail signal d
i(i=1,2, Λ, n).Traditional wavelet transformation denoising method is that the signal after utilizing threshold denoising to individual layer reconstruct is handled, and in the present invention, in order to improve the denoising precision, the FLP algorithm is introduced is into replaced threshold denoising, so that the signal after the reconstruct of each layer individual layer is handled.Owing to can effectively reduce the dispersion degree of the autocorrelation function feature in each frequency range after the multiple dimensioned decomposition of wavelet transformation, therefore will effectively improve the denoising precision of FLP algorithm and improve the denoising speed of convergence, thereby obtain being better than that small echo changes and the denoising result of FLP algorithm.Because the Modeling of Gyro Drift Signal of current time is subjected to the influence of previous moment shifted signal bigger, therefore traditional modeling method utilizes the signal of previous moment to carry out the forward direction one-step prediction as independent variable usually, but, often be difficult to obtain high-precision model because the independent variable dimension is limited.The present invention utilizes neural network that the optical fibre gyro shifted signal is carried out modeling, and itself and wavelet transformation, FLP algorithm merged, each layer approximation signal of the current time of crossing with the FLP algorithm process and detail signal are as the input of neural network, below constantly an optical fibre gyro signal as output network is trained, improve the dimension of input signal, therefore effectively raised modeling accuracy.
The present invention's advantage compared with prior art is:
(1) wavelet transformation can be realized the multiple dimensioned decomposition of signal, thereby has obtained the signal under the different frequency range, has effectively reduced the dispersion degree of the autocorrelation function feature in each frequency range, thereby has improved the denoising precision of FLP algorithm and the speed of convergence of denoising process.
(2) because the advantage of FLP algorithm self makes whole algorithm have characteristics such as real-time, time delay is little, initial procedure is short.
(3) by the reconstruct of wavelet coefficient individual layer, and carry out modeling after respectively each layer coefficients being handled, make that model is accurate more and be easy to the programming realization.
(4) denoising is combined with modeling, effective simplification the processing procedure of optical fibre gyro shifted signal, improved conversion speed.
Description of drawings
Fig. 1 is a FLP-WNN algorithm principle block diagram;
Fig. 2 is the multi-scale wavelet exploded block diagram.
Embodiment
The algorithm principle block diagram of FLP-WNN of the present invention mainly comprises following four steps as shown in Figure 1:
(1): utilize wavelet transformation that the optical fibre gyro signal is carried out multiple dimensioned decomposition
Figure 2 shows that the block diagram that wavelet transformation carries out multiple dimensioned decomposition to the optical fibre gyro signal, wherein S is an original signal, and decomposing the number of plies is 4, approximation signal A4 after obtaining decomposing and detail signal D
i(i=1,2,3,4).Wherein, decompose wavelet basis and elect the db4 small echo as.
(2): to decomposition obtain approach coefficient and detail coefficients is carried out single reconstruct
To decomposition obtain approach coefficient and detail coefficients is carried out single reconstruct, obtain the approximate signal a after the reconstruct
4With detail signal d
i(i=1,2,3,4).Wherein, the reconstruct wavelet basis is elected the db4 small echo as.
(3): the essence that the approximation signal after the reconstruct and detail signal are successively carried out FLP denoising FLP (forward direction linear prediction) algorithm respectively is to utilize in the past the sampled value in N data prediction future.Its main thought is that the gyro signal of previous time output be multiply by the gyro signal that corresponding weights is predicted current time.The acquisition of its optimal weight needs an iterative process, in this process, at first needing to set the weight initial value is zero, calculate the difference between current gyro signal and the predicted value then, minimize this difference according to the LMS least mean square theory, and utilize this difference constantly to adjust and upgrade weight and the final weighted value that obtains a stable convergence.Can utilize the gyro output of previous time to try to achieve the estimated value of current time gyro signal x (n) by following formula:
Wherein, X (n-1)=x (n-1), x (n-2) ..., x (n-N) }
TBe the vector that the output of previous time gyro is formed, x (n-p) is the gyro signal of previous time, α
pBe weight; N is an exponent number.Wherein the exponent number of N is chosen for the key factor that influences this algorithm application, exponent number is big more, and filter effect is good more, but the excessive calculating that can strengthen filtering again of exponent number in addition and influence the real-time of filtering, through repetition test repeatedly, choose N=30 as the exponent number of FLP wave filter herein.Step-length in the FLP wave filter selects to follow following formula:
μ wherein
jBe the step-length under the different frequency range, j=1,2, Λ, n, E
jIt is the average of FLP absolute error in the j frequency range.Approximation signal and detail signal are carried out respectively obtaining a ' after the FLP denoising
4With d '
i(i=1,2,3,4).
(4): network is trained, to obtain the model of zero-drift signal
With the signal a ' after the denoising successively that obtains in the step (3)
4With d '
i(i=1,2,3,4) as the input sample of neural network, below constantly an optical fibre gyro signal as output sample, network is trained, training finishes and promptly obtains the model of zero-drift signal.Wherein output sample is the optical fibre gyro signal based on the current time after the adaptive wavelets transform denoising of forward direction linear prediction (FLP).
In a word, the present invention combines the advantage of wavelet transformation, FLP algorithm and neural network, carry out the multiple dimensioned decomposition of signal by wavelet transformation, effectively reduce the dispersion degree of the autocorrelation function feature in each frequency range, thereby improved the denoising precision of FLP algorithm and the speed of convergence of denoising process.On this basis the wavelet coefficient on each frequency range after the denoising is predicted next optical fibre gyro shifted signal constantly as the input of neural network, so that it is compensated.This method effectively raises denoising and modeling effect, and is significant for the signal accuracy that improves optical fibre gyro.
Claims (7)
1. the Self-adaptation Wavelet Neural Network denoising modeling method based on the forward direction linear prediction is characterized in that comprising the steps:
(1): utilize wavelet transformation that zero-drift signal is carried out multiple dimensioned decomposition
Utilize wavelet transformation that the zero-drift signal under the static state is carried out multiple dimensioned decomposition, the decomposition number of plies is n, obtains decomposing the wavelet coefficient of each layer of back, comprises approximation coefficient and detail coefficients;
(2): to decomposition obtain approach coefficient and detail coefficients is carried out single reconstruct
To decomposition obtain approach coefficient and detail coefficients is carried out single reconstruct, obtain the approximate signal a after the reconstruct
nWith detail signal d
i(i=1,2, Λ, n);
(3): approximation signal after the reconstruct and detail signal are successively carried out the FLP denoising respectively
To the approximation signal a that carries out in the step (2) obtaining after single the reconstruct
nWith detail signal d
i(i=1,2, Λ n) utilizes the FLP algorithm to carry out denoising respectively, obtains signal a (n) and d ' after the denoising
i(i=1,2, Λ, n);
(4): network is trained, to obtain the model of zero-drift signal
With signal a (n) after the denoising successively that obtains in the step (3) and d '
i(i=1,2, Λ, n) as the input sample of neural network, below constantly an optical fibre gyro signal as output sample, network is trained, training finishes and promptly obtains the model of zero-drift signal.
2. according to a kind of Self-adaptation Wavelet Neural Network denoising modeling method according to claim 1, it is characterized in that utilize wavelet transformation that signal is carried out multiple dimensioned decomposition in the described step (1), its wavelet basis is elected the db4 small echo as based on the forward direction linear prediction.
3. a kind of Self-adaptation Wavelet Neural Network denoising modeling method based on the forward direction linear prediction according to claim 1 is characterized in that, pairing approximation signal and detail signal carry out single reconstruct in the described step (2), and its wavelet basis is elected the db4 small echo as.
4. a kind of Self-adaptation Wavelet Neural Network denoising modeling method according to claim 1 based on forward direction linear prediction (FLP), it is characterized in that, utilize the FLP algorithm to carry out denoising in the described step (3), its denoising process is to carry out in the reconstruction of layer of signal, obtain after promptly at first multi-scale wavelet being decomposed approach coefficient and detail coefficients is carried out single reconstruct, then the signal after single the reconstruct is carried out the FLP denoising.
5. a kind of Self-adaptation Wavelet Neural Network denoising modeling method according to claim 1 based on the forward direction linear prediction, it is characterized in that, approximation signal that in the described step (3) multiple dimensioned decomposition is obtained and detail signal successively carry out the FLP denoising respectively, and its FLP wave filter prediction order elects 30 as.
6. a kind of Self-adaptation Wavelet Neural Network denoising modeling method according to claim 1 based on the forward direction linear prediction, it is characterized in that, approximation signal that in the described step (3) multiple dimensioned decomposition is obtained and detail signal successively carry out the FLP denoising respectively, and the step-length in its FLP wave filter selects to follow following formula:
E
j=E[|e
n(n) |]; μ wherein
jBe the step-length under the different frequency range, j=1,2, Λ, n, E
jIt is the average of FLP absolute error in the j frequency range.
7. a kind of Self-adaptation Wavelet Neural Network denoising modeling method according to claim 1 based on the forward direction linear prediction, it is characterized in that, the input node of neural network model is n+1 in the described step (4), and the input sample is a (n) and d ' in the step (3)
i(i=1,2, Λ, n), output node is 1, output sample is the optical fibre gyro signal based on the current time after the adaptive wavelets transform denoising of forward direction linear prediction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011101530251A CN102289715A (en) | 2011-06-08 | 2011-06-08 | Method for adaptively denoising and modeling wavelet neural network based on forward linear prediction (FLP) |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011101530251A CN102289715A (en) | 2011-06-08 | 2011-06-08 | Method for adaptively denoising and modeling wavelet neural network based on forward linear prediction (FLP) |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102289715A true CN102289715A (en) | 2011-12-21 |
Family
ID=45336119
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2011101530251A Pending CN102289715A (en) | 2011-06-08 | 2011-06-08 | Method for adaptively denoising and modeling wavelet neural network based on forward linear prediction (FLP) |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102289715A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103499345A (en) * | 2013-10-15 | 2014-01-08 | 北京航空航天大学 | Fiber-optic gyro temperature drift compensating method based on wavelet analysis and BP (back propagation) neutral network |
| CN103900610A (en) * | 2014-03-28 | 2014-07-02 | 哈尔滨工程大学 | MEMS (Micro-electromechanical Systems) gyroscope random error predication method based on grey wavelet neural network |
| CN104251712A (en) * | 2014-10-09 | 2014-12-31 | 哈尔滨工程大学 | MEMES (micro electro mechanical system) gyroscope random error compensation method on basis of wavelet multi-scale analysis |
| CN106331433A (en) * | 2016-08-25 | 2017-01-11 | 上海交通大学 | Video denoising method based on deep recursive neural network |
| CN108168577A (en) * | 2017-12-22 | 2018-06-15 | 清华大学 | MEMS gyro random error compensation method based on BP neural network |
| CN117076875A (en) * | 2023-10-18 | 2023-11-17 | 中核武汉核电运行技术股份有限公司 | Denoising method for nuclear signal under complex noise background |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101840529A (en) * | 2010-03-26 | 2010-09-22 | 东南大学 | Optic fiber gyroscope random drift modeling method based on locally variable integrated neural network |
-
2011
- 2011-06-08 CN CN2011101530251A patent/CN102289715A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101840529A (en) * | 2010-03-26 | 2010-09-22 | 东南大学 | Optic fiber gyroscope random drift modeling method based on locally variable integrated neural network |
Non-Patent Citations (4)
| Title |
|---|
| 《高技术通讯》 20080229 吉训生等 硅微陀螺信号处理方法--基于前向线性预测小波变换方法 第151页左栏第1段至第154页右栏最后一段，表1 1-7 第18卷, 第2期 * |
| HAIBO HE ET AL: "Home Network Power-Line Communication Signal Processing Based on Wavelet Packet Analysis", 《IEEE TRANSACTIONS ON POWER DELIVERY》 * |
| 吉训生等: "硅微陀螺信号处理方法——基于前向线性预测小波变换方法", 《高技术通讯》 * |
| 陈熙源等: "基于前向线性预测算法的光纤陀螺零漂的神经网络建模", 《中国惯性技术学报》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103499345A (en) * | 2013-10-15 | 2014-01-08 | 北京航空航天大学 | Fiber-optic gyro temperature drift compensating method based on wavelet analysis and BP (back propagation) neutral network |
| CN103499345B (en) * | 2013-10-15 | 2016-04-20 | 北京航空航天大学 | A kind of Fiber Optic Gyroscope Temperature Drift compensation method based on wavelet analysis and BP neural network |
| CN103900610A (en) * | 2014-03-28 | 2014-07-02 | 哈尔滨工程大学 | MEMS (Micro-electromechanical Systems) gyroscope random error predication method based on grey wavelet neural network |
| CN104251712A (en) * | 2014-10-09 | 2014-12-31 | 哈尔滨工程大学 | MEMES (micro electro mechanical system) gyroscope random error compensation method on basis of wavelet multi-scale analysis |
| CN104251712B (en) * | 2014-10-09 | 2017-10-31 | 哈尔滨工程大学 | MEMS gyro random error compensation method based on wavelet multi-scale analysis |
| CN106331433A (en) * | 2016-08-25 | 2017-01-11 | 上海交通大学 | Video denoising method based on deep recursive neural network |
| CN108168577A (en) * | 2017-12-22 | 2018-06-15 | 清华大学 | MEMS gyro random error compensation method based on BP neural network |
| CN117076875A (en) * | 2023-10-18 | 2023-11-17 | 中核武汉核电运行技术股份有限公司 | Denoising method for nuclear signal under complex noise background |
| CN117076875B (en) * | 2023-10-18 | 2024-01-26 | 中核武汉核电运行技术股份有限公司 | Denoising method for nuclear signal under complex noise background |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102289715A (en) | Method for adaptively denoising and modeling wavelet neural network based on forward linear prediction (FLP) | |
| CN109752744B (en) | Multi-satellite combined orbit determination method based on model error compensation | |
| Liu et al. | A deep framework assembling principled modules for CS-MRI: unrolling perspective, convergence behaviors, and practical modeling | |
| CN103557856A (en) | Random drift real-time filtering method for fiber-optic gyroscope | |
| CN110954860B (en) | DOA and polarization parameter estimation method | |
| CN104251712A (en) | MEMES (micro electro mechanical system) gyroscope random error compensation method on basis of wavelet multi-scale analysis | |
| CN110895348A (en) | Method, system and storage medium for extracting low-frequency information of seismic elastic impedance | |
| CN116699526A (en) | Vehicle millimeter wave radar interference suppression method based on sparse and low-rank model | |
| CN108828658A (en) | A kind of ocean bottom seismic data reconstructing method | |
| CN111427096A (en) | Data quality evaluation and filtering processing method for full tensor gravity gradiometer | |
| CN113504505B (en) | One-dimensional DOA estimation method suitable for low signal-to-noise ratio environment | |
| Kimiaefar et al. | Random noise attenuation by Wiener-ANFIS filtering | |
| CN114623848A (en) | Hemispherical resonant gyroscope random error compensation method based on variational modal decomposition and FLP | |
| CN112526599A (en) | Wavelet phase estimation method and system based on weighted L1 norm sparsity criterion | |
| CN106960083A (en) | A kind of robust adaptive beamforming method optimized based on main lobe beam pattern | |
| CN107659290B (en) | Bandwidth extension filter and design method thereof | |
| CN113064134B (en) | Laser radar echo processing method based on improved fuzzy neural network filtering | |
| CN113552565A (en) | Phase unwrapping method for SAR data high-noise and large-gradient change area | |
| CN116299247B (en) | InSAR atmospheric correction method based on sparse convolutional neural network | |
| CN114048636B (en) | Gravity anomaly calculation method and device based on wavelet transformation | |
| CN112231987B (en) | Ionosphere forecasting method based on VMD and Elman neural network | |
| CN114996653A (en) | Two-dimensional robust self-adaptive beam forming method based on atomic norm minimization | |
| CN113702666A (en) | Signal joint noise reduction method for fiber optic gyroscope inertial measurement unit | |
| CN110858309B (en) | Multi-reference time clock weighting synthesis method | |
| CN112800831A (en) | EMD filtering method and system for time-varying gravitational field |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20111221 |