CN118074675B - Adaptive filtering method, device, computer equipment and storage medium - Google Patents
Adaptive filtering method, device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN118074675B CN118074675B CN202410501574.0A CN202410501574A CN118074675B CN 118074675 B CN118074675 B CN 118074675B CN 202410501574 A CN202410501574 A CN 202410501574A CN 118074675 B CN118074675 B CN 118074675B
- Authority
- CN
- China
- Prior art keywords
- signal
- adaptive filter
- neural network
- training
- factor
- 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.)
- Active
Links
Classifications
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03H—IMPEDANCE NETWORKS, e.g. RESONANT CIRCUITS; RESONATORS
- H03H21/00—Adaptive networks
- H03H21/0012—Digital adaptive filters
- H03H21/0043—Adaptive algorithms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03H—IMPEDANCE NETWORKS, e.g. RESONANT CIRCUITS; RESONATORS
- H03H2222/00—Indexing scheme relating to digital filtering methods
- H03H2222/04—Indexing scheme relating to digital filtering methods using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Filters That Use Time-Delay Elements (AREA)
Abstract
The invention relates to the field of signal processing, and discloses a self-adaptive filtering method, a device, computer equipment and a storage medium, wherein the method comprises the following steps: acquiring a step factor after convergence output by a denoising neural network, and configuring the step factor to an adaptive filter; inputting the noise signal into the adaptive filter for filtering to obtain a filtered signal; acquiring a desired signal corresponding to the filtered signal, and calculating an error signal according to the desired signal and the filtered signal; and feeding the error signal back to the adaptive filter, performing adaptive adjustment by the adaptive filter according to the step size factor and the error signal, and then repeatedly executing the step of inputting the noise signal into the adaptive filter for filtering until the filtered signal converges. The step factor is automatically converged by the denoising neural network, so that the adaptive filter can obtain the optimal step factor without manually inputting the step factor.
Description
Technical Field
The present invention relates to the field of signal processing, and in particular, to an adaptive filtering method, apparatus, computer device, and storage medium.
Background
Noise is ubiquitous and in the acquisition of analog and digital signals, whether communication signals or bioelectric acquisition signals (electroencephalogram, electrocardiograph, myoelectricity), there are circuit signals valuable to humans, which must be accompanied by corresponding noise signals. In small signal processing, existing methods of reducing or eliminating noise primarily reduce noise in two ways on the signal collected in the circuit, active noise control (Active Noise Control, ANC) and passive noise control (Passive Noise Control, PNC), respectively. The principle of passive noise is that a physical means is adopted to perform noise elimination to the greatest extent in an acquisition circuit part (such as a digital isolation circuit of an acquisition end and a non-contact acquisition insulating electrode), a Least Mean Square (LMS) algorithm is proposed by Widrow of the university of Steady in 1960, and the LMS algorithm is an adaptive algorithm with small operand, simple algorithm structure and stable algorithm and is widely applied after self-proposal. But it changes the noise filtering speed and accuracy by modifying the algorithm step mu constant, the linear control step factor based method is not always able to solve the contradiction between steady state offset and convergence speed of the adaptive filter. Resulting in a problem of poor processing effect for nonlinear signals.
Disclosure of Invention
In a first aspect, the present application provides an adaptive filtering method, including:
Acquiring a step factor after convergence output by a denoising neural network, and configuring the step factor to an adaptive filter;
inputting the noise signal into the adaptive filter for filtering to obtain a filtered signal;
acquiring a desired signal corresponding to the filtered signal, and calculating an error signal according to the desired signal and the filtered signal;
And feeding the error signal back to the adaptive filter, performing adaptive adjustment by the adaptive filter according to the step size factor and the error signal, and then repeatedly executing the step of inputting the noise signal into the adaptive filter for filtering until the filtered signal converges.
Further, the step factor output by the denoising neural network is obtained, including:
Inputting a training signal into a denoising neural network for training to obtain a noise reduction signal, wherein the training signal comprises a noise signal and an error signal obtained by processing the training signal by the adaptive filter;
inputting the noise reduction signal into a discriminator to obtain a discrimination result, carrying out parameter self-updating on the noise reduction neural network through the discrimination result, and synchronously generating step factors of the self-adaptive filter;
And repeating the step of inputting the training signal into the denoising neural network for training until the step length factor converges, and obtaining the converged step length factor.
Further, the training signal is input into a denoising neural network to perform training, so as to obtain a noise reduction signal, which comprises:
and obtaining a residual signal which is output by the denoising neural network after the training signal is processed, and obtaining the denoising signal by making a difference between the residual signal and the training signal.
Further, the method further comprises:
the authentication result is fed back to the discriminator, and the discriminator updates parameters of the discriminator according to the authentication result.
Further, the authentication result comprises a true value and a false value;
the step of inputting the noise reduction signal into a discriminator to obtain a discrimination result comprises the following steps:
after the noise reduction signal is input into the discriminator, a desired signal corresponding to the noise reduction signal is also input;
and identifying the expected signal and the noise reduction signal, taking the signal identified as a clean signal as the true value, and taking the signal identified as a noise signal as the false value.
Further, the performing parameter self-updating on the denoising neural network according to the identification result includes:
And calculating a loss value of the denoising neural network according to the false value so as to self-update the denoising neural network, and generating a step factor of the adaptive filter when the denoising neural network self-updates.
Further, the feeding back the error signal to the adaptive filter, performing adaptive adjustment according to the step size factor and the error signal, includes:
updating the weighting coefficient of the adaptive filter through a steepest descent method and the step factor, wherein the calculation expression of the weighting coefficient is as follows:
；
Wherein W is the weighting coefficient, n is the nth time, μ is the step factor, e is the error signal, and x is the noise signal.
In a second aspect, the present application provides an adaptive filtering apparatus comprising:
the configuration module is used for acquiring the step factor converged by the denoising neural network and configuring the step factor to the adaptive filter;
The filtering module is used for inputting the noise signal into the adaptive filter for filtering to obtain a filtered signal;
the calculation module is used for acquiring an expected signal corresponding to the filtering signal and calculating an error signal according to the expected signal and the filtering signal;
And the adjusting module is used for feeding the error signal back to the adaptive filter, carrying out adaptive adjustment according to the step factor, and then repeatedly executing the process of inputting the noise signal into the adaptive filter for filtering until the filtered signal converges.
In a third aspect, the application provides a computer device comprising a processor and a memory, the memory storing a computer program which, when run on the processor, performs the adaptive filtering method.
In a fourth aspect, the application provides a readable storage medium storing a computer program which, when run on a processor, performs the adaptive filtering method.
The invention discloses a self-adaptive filtering method, a device, computer equipment and a storage medium, wherein the method comprises the following steps: obtaining a step factor converged by a denoising neural network, and configuring the step factor to an adaptive filter; inputting the noise signal into the adaptive filter for filtering to obtain a filtered signal; acquiring a desired signal corresponding to the filtered signal, and calculating an error signal according to the desired signal and the filtered signal; and feeding the error signal back to the adaptive filter, performing adaptive adjustment according to the step factor, and then repeatedly executing the step of inputting the noise signal into the adaptive filter for filtering until the filtered signal converges. The adaptive filter can obtain the optimal step size factor without manually inputting the step size factor by automatically converging the step size factor through the denoising neural network, so that the parameter variable of the adaptive filter can be more accurately determined, and the contradiction between the steady state offset and the convergence speed is improved.
Drawings
In order to more clearly illustrate the technical solutions of the present invention, the drawings that are required for the embodiments will be briefly described, it being understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope of the present invention. Like elements are numbered alike in the various figures.
FIG. 1 is a schematic flow chart of an adaptive filtering method according to an embodiment of the present application;
FIG. 2 is a schematic diagram of a step factor convergence procedure according to an embodiment of the present application;
FIG. 3 is a schematic diagram of a DnCNN model according to an example embodiment of the present application;
FIG. 4 is a schematic diagram showing a step factor convergence according to a real-time example of the present application;
fig. 5 shows a schematic structural diagram of an adaptive filtering device according to an embodiment of the present application.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments.
The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by a person skilled in the art without making any inventive effort, are intended to be within the scope of the present invention.
The terms "comprises," "comprising," "including," or any other variation thereof, are intended to cover a specific feature, number, step, operation, element, component, or combination of the foregoing, which may be used in various embodiments of the present invention, and are not intended to first exclude the presence of or increase the likelihood of one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing.
Furthermore, the terms "first," "second," "third," and the like are used merely to distinguish between descriptions and should not be construed as indicating or implying relative importance.
Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which various embodiments of the invention belong. The terms (such as those defined in commonly used dictionaries) will be interpreted as having a meaning that is the same as the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein in connection with the various embodiments of the invention.
The technical scheme of the application discloses a self-adaptive filtering method, in particular to a self-adaptive filtering method utilizing a denoising neural network model, which carries out cyclic iteration through the self-updating characteristic of the denoising neural network model to generate a converged step factor, so that the self-adaptive filter can carry out rapid and accurate self-adaptive adjustment through the step factor, and the value of the step factor is more scientific and practical.
The technical scheme of the application is described in the following specific embodiments.
Example 1
As shown in fig. 1, the adaptive filtering method of the present embodiment includes:
step S100, obtaining a step factor after convergence output by the denoising neural network, and configuring the step factor to the adaptive filter.
In this embodiment, the adaptive filter is an LMS (least mean square error) adaptive filter, and when the filter works, the adaptive filter performs adaptive adjustment according to the output and the expected value, so that the output result converges to obtain a filtered signal that is close to the expected value. The step factor is a parameter that affects its adaptive adjustment. The present embodiment automatically acquires the step factor by denoising the neural network, not manually set.
Specifically, the flow of obtaining the step factor is shown in fig. 2:
Step S110, a training signal is input into a denoising neural network to perform training, so as to obtain a noise reduction signal, wherein the training signal comprises a noise signal and an error signal obtained by processing the training signal by the adaptive filter.
The training signal is a signal input to the denoising neural network for obtaining a step factor, and is composed of a noise signal and an error signal obtained by processing the training signal by the adaptive filter.
The denoising neural network of the present embodiment may be, for example, a model of an anti-denoising convolutional neural network (Denoising Convolutional Neural Network, dnCNN).
For a better explanation of the operation of the process, reference is made to the workflow within the DnCNN model as shown in fig. 3.
When the noise-containing signal is input DnCNN, a residual signal can be output, and a corresponding noise-reducing signal can be obtained by making a difference between the residual signal and the noise signal input before. The noise-containing signal in this embodiment is a training signal.
And step S120, inputting the noise reduction signal into a discriminator to obtain a discrimination result, carrying out parameter self-updating on the noise reduction neural network through the discrimination result, and synchronously generating step factors of the self-adaptive filter.
The discriminator is a module for discriminating whether the input signal is a noisy signal or a filtered signal, the discrimination result includes a true value and a false value, and after the noise reduction signal is input to the discriminator, a desired signal corresponding to the noise reduction signal is also input. And identifying the expected signal and the noise reduction signal, taking the signal identified as a clean signal as the true value, and taking the signal identified as a noise signal as the false value.
The desired signal is the original clean signal.
And calculating a loss value of the denoising neural network according to the false value so as to self-update the denoising neural network, and generating a step factor of the adaptive filter when the denoising neural network self-updates.
It can be seen that the neural network in combination with the adaptive filtering algorithm actually builds a learned network structure, and searches for the optimal learning step size through the error and the input signal. And learning given sample data, establishing a real-time data association model, integrating the model into DnCNN algorithm, and finally outputting step factors while self-updating.
In addition, both the true and false values are returned to the discriminator to update the parameters of the discriminator to increase the accuracy of the discriminator.
And step S130, repeatedly executing the step of inputting the training signal into the denoising neural network for training until the step factor is converged, and obtaining the converged step factor.
Since DnCNN is a self-updatable model, it can perform u-phase iteration, thus keeping the input electrical signal unchanged, repeating the operations from step S110 to step S130, and further improving the output step size factor, and as DnCNN is iteratively updated, the parameters of DnCNN model itself will converge, and the output step size factor will also converge, so that when the step size factor converges, the iteration can be ended, and the output step size factor is used as the final output on the LMS filter.
Specifically, as shown in fig. 4, the whole process is related to the step factor acquisition.
The noise signal is input to the DnCNN model and the adaptive filter, which will obtain a filtered signal, which can then be calculated in combination with the desired signal to obtain an error signal, which is input to the DnCNN model, and the DnCNN model outputs a step factor with the workflow of fig. 3. And then repeatedly acquiring a training signal consisting of the noise signal and the error signal, so as to perform iterative self-updating operation and realize convergence of the step factor. In this process, the output error signal carries the characteristics of the adaptive filter, so that the DnCNN model can output a step size factor suitable for the adaptive filter over multiple iteration cycles.
It should be noted that, when step S110 is repeatedly performed, the noise signal is unchanged, and the adaptive filter may change the error signal due to its own characteristics, so that after the network training is finished, the parameters obtained by the DnCNN neural network algorithm represent the nonlinear correlation among the input signal, the error and the step factor. This results in a final step factor that has both convergence speed and convergence accuracy for the adaptive filter.
Moreover, the loop in fig. 4 only represents the process of calculating DnCNN the step size factor of the convergence, but not the process when the adaptive filter actually performs the filtering operation, when the step size factor of the convergence is obtained, the DnCNN model does not perform the training operation, and the adaptive filter keeps the step size factor unchanged to perform the actual adaptive filtering operation.
Step S200, inputting the noise signal into the adaptive filter for filtering to obtain a filtered signal.
In the actual filtering operation, after receiving the noise signal, the LMS filter outputs a filtered signal according to the current weighting coefficient, specifically, the expression of the filter is:
y（n）=WT（n）x（n）；
Where W is a weighting coefficient of the filter, W (n) = [ W 0,w1,… wL-1]T ], L is an order of the filter, n is an nth time, x is a noise signal, and y is an output filtered signal.
It can be seen that the weighting coefficients are correlated with the final output filtered signal, and the adaptive filter can automatically calculate an error signal from the output filtered signal, adjusting the weighting coefficients based on the error signal, resulting in a more accurate output. The precise and rapid adjustment of the weighting coefficients is thus the adjustment of the filtered signal.
Step S300, obtaining a desired signal corresponding to the filtered signal, and calculating an error signal according to the desired signal and the filtered signal.
After the filtered signal is obtained, a corresponding expected signal can be obtained, wherein the expected signal is an original clean signal without noise, and an error signal can be obtained by making a difference between the two signals, and the error signal comprises noise which is not filtered and some signal data which is not filtered. The specific expression is as follows:
e（n）=d（n）-y（n）；
where e is the error signal and d is the desired signal. The larger the error is, the worse the filtering effect is, the smaller the error is, and the better the filtering effect is.
Taking the mean square value of the error signal as the optimal statistical criterion, a cost function F (n) =e (E 2 (n)) can be defined.
The change of F along with time characterizes the convergence speed and convergence precision of the filter, and when F takes the minimum value, the weighting coefficient of the filter reaches the optimum.
And step 400, feeding back the error signal to the adaptive filter, performing adaptive adjustment by the adaptive filter according to the step factor and the error signal, and then repeatedly executing the step of inputting the noise signal into the adaptive filter for filtering until the filtered signal converges.
To obtain the optimal weighting coefficients, LMS filters are typically tuned using the steepest descent method, the weighting coefficients are calculated as:
；
wherein μ is the step factor.
It can be seen from the equation that each cycle e (n) becomes smaller, W (n) and W (n+1) become closer and closer to each other until convergence to a limit, and the step factor μ controls the rate of change of the weighting coefficient, so that the larger the step, the faster the convergence rate, but the worse the convergence accuracy, and vice versa. Thus, after a certain number of cycles, it is possible to determine that the filtered signal converges and output the filtered signal.
The step size factor is usually manually set, and is not changed after setting, but the step size factor in this embodiment is generated by the steps S110 to S130, which has a faster convergence speed and higher accuracy than the conventional standard LMS filter.
According to the embodiment, self-adaptive learning is performed through the DnCNN model, step factors of the self-adaptive filter are generated, so that the step factors of the self-adaptive filter can be automatically generated without manual input, the DnCNN model can construct three-side relations among noise signals, error signals and the step factors, the obtained step factors can deal with linear signal processing scenes and nonlinear signal processing scenes, and the combination of the self-adaptive filter and the LMS self-adaptive filter is realized, so that parameter variables of the LMS filter are more accurately determined, and contradiction between steady state offset and convergence speed is improved.
Example 2
As shown in fig. 5, this embodiment further provides an adaptive filtering apparatus, including:
A configuration module 10, configured to obtain a step factor after convergence output by the denoising neural network, and configure the step factor to the adaptive filter;
The filtering module 20 is configured to input a noise signal into the adaptive filter for filtering, so as to obtain a filtered signal;
A calculating module 30, configured to obtain a desired signal corresponding to the filtered signal, and calculate an error signal according to the desired signal and the filtered signal;
And the adjusting module 40 is configured to feed back the error signal to the adaptive filter, and the adaptive filter performs adaptive adjustment according to the step size factor and the error signal, and then repeatedly perform the filtering by inputting a noise signal into the adaptive filter until the filtered signal converges.
The application provides a computer device comprising a processor and a memory, said memory storing a computer program which, when run on said processor, performs said adaptive filtering method.
The present application provides a readable storage medium storing a computer program which when run on a processor performs the adaptive filtering method.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. The apparatus embodiments described above are merely illustrative, for example, of the flow diagrams and block diagrams in the figures, which illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
In addition, functional modules or units in various embodiments of the invention may be integrated together to form a single part, or the modules may exist alone, or two or more modules may be integrated to form a single part.
The functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a smart phone, a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a usb disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
The foregoing is merely illustrative of the present invention, and the present invention is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present invention.
Claims (8)
1. An adaptive filtering method, comprising:
Acquiring a step factor after convergence output by a denoising neural network, and configuring the step factor to an adaptive filter;
inputting the noise signal into the adaptive filter for filtering to obtain a filtered signal;
acquiring a desired signal corresponding to the filtered signal, and calculating an error signal according to the desired signal and the filtered signal;
feeding back the error signal to the adaptive filter, performing adaptive adjustment by the adaptive filter according to the step size factor and the error signal, and then repeatedly executing the step of inputting the noise signal into the adaptive filter for filtering until the filtered signal converges;
the step factor output by the denoising neural network is obtained, which comprises the following steps:
Inputting a training signal into a denoising neural network for training to obtain a noise reduction signal, wherein the training signal comprises a noise signal and an error signal obtained by processing the training signal by the adaptive filter;
inputting the noise reduction signal into a discriminator to obtain a discrimination result, carrying out parameter self-updating on the noise reduction neural network through the discrimination result, and synchronously generating step factors of the self-adaptive filter;
repeating the step of inputting the training signal into the denoising neural network for training until the step length factor converges, and obtaining the converged step length factor;
the training signal is input into a denoising neural network for training to obtain a denoising signal, which comprises the following steps:
and obtaining a residual signal which is output by the denoising neural network after the training signal is processed, and obtaining the denoising signal by making a difference between the residual signal and the training signal.
2. The adaptive filtering method according to claim 1, further comprising:
the authentication result is fed back to the discriminator, and the discriminator updates parameters of the discriminator according to the authentication result.
3. The adaptive filtering method according to claim 1, wherein the discrimination result includes a true value and a false value;
the step of inputting the noise reduction signal into a discriminator to obtain a discrimination result comprises the following steps:
after the noise reduction signal is input into the discriminator, a desired signal corresponding to the noise reduction signal is also input;
and identifying the expected signal and the noise reduction signal, taking the signal identified as a clean signal as the true value, and taking the signal identified as a noise signal as the false value.
4. The adaptive filtering method according to claim 3, wherein the performing parameter self-updating on the denoising neural network according to the discrimination result comprises:
And calculating a loss value of the denoising neural network according to the false value so as to self-update the denoising neural network, and generating a step factor of the adaptive filter when the denoising neural network self-updates.
5. The adaptive filtering method of claim 1, wherein said feeding back the error signal to the adaptive filter, adaptively adjusting according to the step size factor and the error signal, comprises:
updating the weighting coefficient of the adaptive filter through a steepest descent method and the step factor, wherein the calculation expression of the weighting coefficient is as follows:
；
Wherein W is the weighting coefficient, n is the nth time, μ is the step factor, e is the error signal, and x is the noise signal.
6. An adaptive filtering device, comprising:
The configuration module is used for acquiring the converged step size factor output by the denoising neural network and configuring the step size factor to the adaptive filter;
The filtering module is used for inputting the noise signal into the adaptive filter for filtering to obtain a filtered signal;
The calculation module is used for acquiring an expected signal corresponding to the filtering signal and calculating an error signal according to the expected signal and the filtering signal;
The adjusting module is used for feeding back the error signal to the adaptive filter, the adaptive filter carries out adaptive adjustment according to the step size factor and the error signal, and then the noise signal is repeatedly input into the adaptive filter for filtering until the filtering signal converges;
the step factor output by the denoising neural network is obtained, which comprises the following steps:
Inputting a training signal into a denoising neural network for training to obtain a noise reduction signal, wherein the training signal comprises a noise signal and an error signal obtained by processing the training signal by the adaptive filter;
inputting the noise reduction signal into a discriminator to obtain a discrimination result, carrying out parameter self-updating on the noise reduction neural network through the discrimination result, and synchronously generating step factors of the self-adaptive filter;
repeating the step of inputting the training signal into the denoising neural network for training until the step length factor converges, and obtaining the converged step length factor;
the training signal is input into a denoising neural network for training to obtain a denoising signal, which comprises the following steps:
and obtaining a residual signal which is output by the denoising neural network after the training signal is processed, and obtaining the denoising signal by making a difference between the residual signal and the training signal.
7. A computer device comprising a processor and a memory, the memory storing a computer program that, when run on the processor, performs the adaptive filtering method of any one of claims 1 to 5.
8. A readable storage medium, characterized in that it stores a computer program which, when run on a processor, performs the adaptive filtering method of any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410501574.0A CN118074675B (en) | 2024-04-25 | 2024-04-25 | Adaptive filtering method, device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410501574.0A CN118074675B (en) | 2024-04-25 | 2024-04-25 | Adaptive filtering method, device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118074675A CN118074675A (en) | 2024-05-24 |
| CN118074675B true CN118074675B (en) | 2024-07-02 |
Family
ID=91099458
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410501574.0A Active CN118074675B (en) | 2024-04-25 | 2024-04-25 | Adaptive filtering method, device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118074675B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107230473A (en) * | 2017-06-30 | 2017-10-03 | 邢优胜 | Noise initiative control method and system in a kind of submarine cabin |
| CN109714023A (en) * | 2018-12-28 | 2019-05-03 | 歌尔股份有限公司 | Adaptive filter method, sef-adapting filter and noise control system |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3147864B2 (en) * | 1998-07-16 | 2001-03-19 | 日本電気株式会社 | Adaptive step size control adaptive filter and adaptive step size control method |
| US9296474B1 (en) * | 2012-08-06 | 2016-03-29 | The United States of America as represented by the Administrator of the National Aeronautics & Space Administration (NASA) | Control systems with normalized and covariance adaptation by optimal control modification |
| CN110113031A (en) * | 2019-04-18 | 2019-08-09 | 五邑大学 | An adaptive filtering method, device and storage medium thereof |
| CN117454102A (en) * | 2023-10-16 | 2024-01-26 | 浙江工业大学 | Self-adaptive noise elimination method and device for river-crossing pipeline positioning detection system based on FPGA |
-
2024
- 2024-04-25 CN CN202410501574.0A patent/CN118074675B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107230473A (en) * | 2017-06-30 | 2017-10-03 | 邢优胜 | Noise initiative control method and system in a kind of submarine cabin |
| CN109714023A (en) * | 2018-12-28 | 2019-05-03 | 歌尔股份有限公司 | Adaptive filter method, sef-adapting filter and noise control system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118074675A (en) | 2024-05-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mathews et al. | A stochastic gradient adaptive filter with gradient adaptive step size | |
| Arikan et al. | Adaptive filtering for non-Gaussian stable processes | |
| EP0086943B1 (en) | Adaptive spectrum shaping filter | |
| WO2007053831A2 (en) | Optimum nonlinear correntropy filter | |
| Nascimento et al. | Probability of divergence for the least-mean fourth algorithm | |
| Lim et al. | A piloted adaptive notch filter | |
| CN109379652B (en) | Earphone active noise control secondary channel off-line identification method | |
| CN117997312A (en) | Variable step length self-adaptive filtering method and system | |
| CN118074675B (en) | Adaptive filtering method, device, computer equipment and storage medium | |
| Cousseau et al. | Factorized all-pass based IIR adaptive notch filters | |
| WO2024072700A1 (en) | Switchable noise reduction profiles | |
| CN112309362A (en) | Active acoustic noise reduction method and device of MR system and computer equipment | |
| Kumari et al. | Performance Evaluation of Adaptive Filtering Algorithms for Denoising the ECG Signal | |
| CN113299284B (en) | Speech recognition device, method, equipment and medium based on adaptive filtering | |
| LeBlanc et al. | A two-stage deep neuroevolutionary technique for self-adaptive speech enhancement | |
| CN112420066A (en) | Noise reduction method, noise reduction device, computer equipment and computer readable storage medium | |
| Vaishnavi et al. | Noise removal from ECG signal using LMS adaptive filter implementation in Xilinx system generator | |
| CN100589184C (en) | Subtractive cancellation of harmonic noise | |
| Lim et al. | Adaptive algorithms for joint time delay estimation and IIR filtering | |
| CN112395946A (en) | Mud pulse signal denoising method based on wavelet transform-RBF neural network | |
| Bergamasco et al. | Active noise control of impulsive noise with selective outlier elimination | |
| CN118100914B (en) | Dual-frequency synthesizer control method and system based on artificial intelligence | |
| Kelly et al. | Frequency tracking and variable bandwidth for line noise filtering without a reference | |
| Fathey et al. | Enhancement of ECG signal | |
| Sucher | A self-organizing nonlinear filter based on fuzzy clustering |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |