Signal noise suppression and signal integrity guaranteeing method based on deep learning

Abstract

The invention discloses a deep learning-based signal noise suppression signal integrity guaranteeing method, and relates to the technical field of noise suppression processing, and the suppression processing steps are as follows: obtaining noise, analyzing the noise based on a spectral analysis method, performing comparative analysis on the noise type based on big data analysis, and judging the type of the current noise; constructing a neural network model, training and optimizing the model, setting an adaptive learning mechanism, and analyzing different noise types; the extracted noise signals are correspondingly input into the constructed model, and the integrity of the signals is guaranteed while a noise source is suppressed. According to the method, the source of noise is analyzed and judged, different processing means are adopted for different types of noise, the neural network model is optimized, a self-adaptive learning mechanism is set, classified noise signals are suppressed, the smoothing effect during noise processing is improved, and after noise is suppressed, the noise quality is improved. And the integrity and the original characteristics of the signal can be better ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of noise suppression processing, and in particular to a method for suppressing signal noise based on deep learning to ensure signal integrity. Background Art

[0002] Noise is a widely used concept in many fields such as signal processing, communication, acoustics, and electronic engineering. It usually refers to all unwanted, interfering signals or fluctuations outside of useful signals. The sources of noise are varied. It can be natural, such as wind, rain, electromagnetic radiation, etc., or artificial, such as machine noise, electromagnetic interference of electronic equipment, etc. In signal processing, noise is usually regarded as a random signal, and its characteristics may include random changes in amplitude, frequency, phase, etc. The presence of noise will interfere with or mask useful signals, making the reception, recognition, and analysis of signals difficult. Therefore, in signal processing, noise suppression is an important task; the existing method of suppressing sound waves is to remove the noise components in the signal by filtering, but the filtering method often does not have an ideal smoothing effect on the noise, and after suppressing the noise, the integrity of the signal cannot be guaranteed, resulting in the original characteristics of the signal and affecting the quality of the signal. In this regard, we propose a signal noise suppression method based on deep learning to ensure signal integrity. Summary of the invention

[0003] In order to solve the above technical problems, a signal noise suppression method based on deep learning to ensure signal integrity is provided. This technical solution solves the above problems.

[0004] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0005] A signal noise suppression method based on deep learning to ensure signal integrity. The suppression processing steps are as follows:

[0006] Acquire noise, analyze it based on spectrum analysis, and compare and analyze noise types based on big data analysis to determine the type of current noise;

[0007] Build a neural network model, train and optimize the model, set up an adaptive learning mechanism, and analyze different noise types;

[0008] The extracted noise signal is input into the constructed model accordingly to suppress the noise source while ensuring the integrity of the signal.

[0009] Preferably, noise acquisition is performed based on collection and acquisition of audio acquisition equipment, and the acquired noise is uploaded to a computer for processing; the spectrum analysis method samples the continuous-time signal in advance, performs Fourier transform processing on the sampled discrete-time signal, and then calculates the spectrum to calculate the amplitude and phase of the signal in different frequency domains, and plots the calculated spectrum to represent it in the form of a graph.

[0010] Preferably, the calculation formula for signal sampling is:

[0011] x(n)=x(nT s )

[0012] Where T s is the sampling period, x(n) is the discrete time signal, and n is an integer representing the index of the sampling point;

[0013] The Fourier transform formula is:

[0014]

[0015] Where N is the total number of sample points, k is the frequency index, the value range is 0≤k≤N-1, j is the imaginary unit, X(k) is the frequency domain, e is the base of the natural logarithm, and the amplitude and phase of the signal in different frequency domains are calculated from the results obtained by Fourier transform.

[0016] Preferably, multiple groups of noise spectrum graphs are obtained based on big data, and the multiple groups of noise spectrum graphs are compared and calculated with the current noise spectrum graph to determine whether they are consistent. The calculation formula is:

[0017]

[0018] Where o is the number of samples, X i With Y i are the observed values ​​of the two variables at the sample point i on the spectrum graph, and are the sample means of X and Y, r is the similarity of the two sets of spectrograms, where r value of 1 indicates complete correlation, that is, the two sets of spectrograms are consistent, and other values ​​indicate inconsistency. The closer the value of r is to 1, the more consistent the two sets of spectrograms are, and vice versa. The calculation is used to determine whether the currently collected noise signal is consistent with the noise spectrum obtained in the big data to obtain the type of the currently collected noise signal.

[0019] Preferably, the noise signal types include natural noise and man-made noise, wherein natural noise includes electrical noise and atmospheric noise. Natural noise originates from noise generated in the natural environment, and man-made noise is generated based on man-made activities, including industrial noise, mechanical noise and electronic product noise. The noise in the natural environment and the noise signal categories generated by humans are obtained through big data to obtain corresponding noise signals, and the type of current noise can be determined by comparing them with the noise signals obtained from big data.

[0020] Preferably, the model training method is based on the gradient descent method to update the weights and bias terms, and the model is optimized based on the regularization method to prevent overfitting. The Adam algorithm is selected as the adaptive learning mechanism, the algorithm parameters are configured in the model, and iterative optimization processing is continuously performed to improve the performance of the neural network model.

[0021] Preferably, when the currently classified noise signal is input, the neural network model is provided with a dual input framework, and different corresponding noise signals are input into corresponding input entrances in the neural network model, wherein the dual input framework of the neural network model includes a natural noise input entrance and an artificial noise input entrance, and the corresponding input is input into different input entrances after judging the currently input noise.

[0022] Preferably, when training the model, the parameters and structure of the model are dynamically adjusted according to the performance of the model on the validation set. When training the model, two sets of training sets are prepared, one set is an artificial noise data set, and the other set is a natural noise data set. Each set of data sets is preprocessed. The model contains two parallel sub-networks, one for processing artificial noise data and the other for processing natural noise data.

[0023] Preferably, the iterative training formula is:

[0024]

[0025] Among them, θ1, θ2 and θ n is the parameter of the model, L(θ) is the loss function, and the parameters are updated. The update formula is:

[0026]

[0027] Where η is the learning rate. In each iteration of the model, the step parameters are updated according to the formula to optimize the performance of the neural network model.

[0028] Preferably, the extracted noise signal is input into the corresponding input port of the trained model, and the model suppresses the noise according to the learned knowledge. While suppressing the noise, the model ensures that the integrity of the signal is not affected and retains the useful information in the signal, such as frequency components and phase information.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] The method proposed in the present invention analyzes and determines the source of noise, adopts different processing methods for different types of noise, optimizes the neural network model, and sets up an adaptive learning mechanism to suppress the classified noise signal, improve the smoothing effect during noise processing, and better protect the integrity and original characteristics of the signal after suppressing the noise. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a flowchart of the noise suppression processing steps of the present invention;

[0032] Figure 2 This is a flow chart of noise type judgment and analysis of the present invention. DETAILED DESCRIPTION

[0033] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0034] Reference Figure 1 As shown in the figure, the signal noise suppression method based on deep learning to ensure signal integrity, the suppression processing steps are:

[0035] Acquire noise, analyze it based on spectrum analysis, and compare and analyze noise types based on big data analysis to determine the type of current noise;

[0036] Build a neural network model, train and optimize the model, set up an adaptive learning mechanism, and analyze different noise types;

[0037] The extracted noise signal is input into the constructed model accordingly to suppress the noise source while ensuring the integrity of the signal.

[0038] By performing spectral analysis on noise signals, the frequency distribution characteristics of noise signals can be clearly displayed. This method can reveal the spectral characteristics of noise and help distinguish different types of noise. Based on big data analysis technology, a large amount of noise data can be deeply mined to identify the source, distribution and changing trend of noise. By comparing and analyzing noise data from different sources, the type of current noise can be more accurately determined, providing a strong basis for subsequent noise control; construct and train a neural network model so that it can learn and identify different types of noise signals. The neural network model has strong pattern recognition capabilities and can adaptively adjust parameters to cope with complex noise environments. An adaptive learning mechanism is set up so that the neural network model can continuously optimize its performance in practical applications. When encountering new noise types or changes in the noise environment, the model can automatically adjust its parameters and strategies to better suppress the noise source. The application of noise spectrum analysis, big data analysis and neural network models can promote the continuous innovation and development of noise control technology.

[0039] Reference Figure 2 As shown, noise acquisition is based on the collection and acquisition of audio acquisition equipment, and the acquired noise is uploaded to the computer for processing; the spectrum analysis method samples the continuous time signal in advance, performs Fourier transform processing on the sampled discrete time signal, and then calculates the spectrum to calculate the amplitude and phase of the signal in different frequency domains, and draws the calculated spectrum to represent it in the form of a graph.

[0040] The calculation formula for signal sampling is:

[0041] x(n)=x(nT s )

[0042] Where T s is the sampling period, x(n) is the discrete time signal, and n is an integer representing the index of the sampling point;

[0043] The Fourier transform formula is:

[0044]

[0045] Where N is the total number of sample points, k is the frequency index, the value range is 0≤k≤N-1, j is the imaginary unit, X(k) is the frequency domain, e is the base of the natural logarithm, and the amplitude and phase of the signal in different frequency domains are calculated from the results obtained by Fourier transform.

[0046] Based on big data, multiple groups of noise spectra are obtained, and the noise spectra of multiple groups are compared and calculated with the current noise spectra to determine whether they are consistent. The calculation formula is:

[0047]

[0048] Where o is the number of samples, X i With Y i are the observed values ​​of the two variables at the sample point i on the spectrum graph, and are the sample means of X and Y, r is the similarity of the two sets of spectrograms, where r value of 1 indicates complete correlation, that is, the two sets of spectrograms are consistent, and other values ​​indicate inconsistency. The closer the value of r is to 1, the more consistent the two sets of spectrograms are, and vice versa. The calculation is used to determine whether the currently collected noise signal is consistent with the noise spectrum obtained in the big data to obtain the type of the currently collected noise signal.

[0049] Spectrum calculation is an existing technology, and no further elaboration is given. The spectrum drawing method is based on using MATLAB drawing tools to draw the spectrum graph. The drawing steps are as follows:

[0050] Set drawing parameters;

[0051] The horizontal axis represents the frequency in Hz or other appropriate frequency units;

[0052] The vertical axis represents amplitude and power. In a logarithmic scale, decibels (dB) are usually used as units to more clearly show the relative sizes of different frequency components;

[0053] Add grid lines as needed to allow for more accurate readings;

[0054] Add a title, horizontal and vertical axis labels, and necessary legends to the spectrum graph;

[0055] Draw a spectrum diagram based on the calculated spectrum data and the set drawing parameters. The spectrum diagram can show the distribution of the signal at different frequencies, thus helping to analyze the frequency domain characteristics of the signal.

[0056] The spectrum diagram records and plots the graphical data of the signal at various frequencies in the form of ripples on the horizontal and vertical axes. It is a commonly used analysis tool in the field of signal processing and communication.

[0057] Noise signal types include natural noise and man-made noise. Natural noise includes electrical noise and atmospheric noise. Natural noise originates from noise generated in the natural environment. Man-made noise is generated based on human activities. Man-made noise includes industrial noise, mechanical noise and electronic product noise. The noise in the natural environment and the noise signal categories generated by humans are obtained through big data to obtain corresponding noise signals. By comparing them with the noise signals obtained from big data, the type of current noise can be determined.

[0058] The model training method is based on the gradient descent method to update the weights and bias terms, and optimizes the model based on the regularization method to prevent overfitting. The Adam algorithm is selected as the adaptive learning mechanism, the algorithm parameters are configured in the model, and iterative optimization is continuously performed to improve the performance of the neural network model.

[0059] The gradient descent method and regularization method in this application are prior arts, and will not be elaborated here. The core idea of ​​the Adam algorithm is to estimate the learning rate by using the mean of the first-order moment and the variance of the second-order moment of the gradient, and adaptively update the model parameters during the training process, and maintain two dynamically adjusted learning rates, one is the momentum term based on the estimation of the first-order moment of the gradient, which is used to smooth the direction of the gradient update; the other is the adaptive learning rate based on the estimation of the second-order moment of the gradient, which is used to dynamically adjust the size of the learning rate; the Adam algorithm can adaptively adjust the learning rate according to the gradient characteristics of different parameters. For parameters with larger gradients, the learning rate will be reduced accordingly to avoid oscillation caused by too fast parameter updates; for parameters with smaller gradients, the learning rate will be increased accordingly to accelerate convergence. By combining the momentum method, the Adam algorithm obtains a more stable and consistent update direction in the parameter space, thereby having a faster convergence speed. The Adam algorithm can better adapt to sparse gradients through adaptability and reduce oscillation during training. Due to the introduction of momentum and second-order moment estimation, the Adam algorithm has a certain anti-interference ability to noise during training, making the model more stable.

[0060] When the currently classified noise signal is input, the neural network model is provided with a dual input framework, and different corresponding noise signals are input into the corresponding input entrances of the neural network model, wherein the dual input framework of the neural network model includes a natural noise input entrance and an artificial noise input entrance, and the corresponding input is input into different input entrances after judging the noise of the current input.

[0061] In this application, by setting input entries for natural noise and artificial noise respectively, the model can process these two different types of noise more specifically. Each type of noise has its own unique characteristics and influences. Separate processing can enable the model to better learn and identify the characteristics of each noise, thereby improving the processing effect. Inputting different types of noise separately can reduce mutual interference between them. The dual-input framework allows the model to learn natural noise and artificial noise independently. This learning method helps the model to have a deeper understanding of the nature and changing laws of each type of noise, thereby enhancing the generalization ability of the model. Even when encountering new types of noise, the model can make inferences and judgments based on existing learning experience.

[0062] When training the model, the parameters and structure of the model are dynamically adjusted according to the performance of the model on the validation set. When training the model, two sets of training sets are prepared, one is an artificial noise data set and the other is a natural noise data set. Each data set is preprocessed. The model contains two parallel sub-networks, one for processing artificial noise data and the other for processing natural noise data.

[0063] The iterative training formula is:

[0064]

[0065] Among them, θ1, θ2 and θ n is the parameter of the model, L(θ) is the loss function, and the parameters are updated. The update formula is:

[0066]

[0067] Where η is the learning rate. In each iteration of the model, the step parameters are updated according to the formula to optimize the performance of the neural network model.

[0068] In this application, iterative training allows the model to gradually learn and optimize over multiple iterations. Each iteration is adjusted based on the results of the previous iteration, so that the model can more accurately capture the patterns and rules in the data. Through iterative training, the complexity and parameters of the model can be gradually adjusted to reduce the risk of overfitting. Overfitting is a phenomenon in which the model performs well on training data but poorly on new data. Techniques such as cross-validation in iterative training can help identify and prevent overfitting.

[0069] The extracted noise signal is input into the corresponding input port of the trained model. The model suppresses the noise based on the learned knowledge. While suppressing the noise, the model ensures that the integrity of the signal is not affected and retains the useful information in the signal, such as frequency components and phase information.

[0070] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A signal noise suppression method based on deep learning to ensure signal integrity, characterized in that: The inhibition process steps are: Acquire noise, analyze it based on spectrum analysis, and compare and analyze noise types based on big data analysis to determine the type of current noise; Build a neural network model, train and optimize the model, set up an adaptive learning mechanism, and analyze different noise types; The extracted noise signal is input into the constructed model accordingly to suppress the noise source while ensuring the integrity of the signal.

2. The method for signal noise suppression and signal integrity protection based on deep learning according to claim 1, characterized in that: Noise acquisition is based on the collection and acquisition of audio acquisition equipment, and the acquired noise is uploaded to the computer for processing; the spectrum analysis method samples the continuous time signal in advance, performs Fourier transform processing on the sampled discrete time signal, and then calculates the spectrum to calculate the amplitude and phase of the signal in different frequency domains, and draws the calculated spectrum to express it in the form of a graph.

3. The method for signal noise suppression and signal integrity protection based on deep learning according to claim 2, characterized in that: The calculation formula for signal sampling is: x(n)=x(nT s ) Where T s is the sampling period, x(n) is the discrete time signal, and n is an integer representing the index of the sampling point; The Fourier transform formula is: Where N is the total number of sample points, k is the frequency index, the value range is 0≤k≤N-1, j is the imaginary unit, X(k) is the frequency domain, e is the base of the natural logarithm, and the amplitude and phase of the signal in different frequency domains are calculated from the results obtained by Fourier transform.

4. The method for signal noise suppression and signal integrity protection based on deep learning according to claim 1, characterized in that: Based on big data, multiple groups of noise spectra are obtained, and the noise spectra of multiple groups are compared and calculated with the current noise spectra to determine whether they are consistent. The calculation formula is: Where o is the number of samples, X i With Y i are the observed values ​​of the two variables at the sample point i on the spectrum graph, and are the sample means of X and Y, r is the similarity of the two sets of spectrograms, where r value of 1 indicates complete correlation, that is, the two sets of spectrograms are consistent, and other values ​​indicate inconsistency. The closer the value of r is to 1, the more consistent the two sets of spectrograms are, and vice versa. The calculation is used to determine whether the currently collected noise signal is consistent with the noise spectrum obtained in the big data to obtain the type of the currently collected noise signal.

5. The method for signal noise suppression and signal integrity protection based on deep learning according to claim 1, characterized in that: Noise signal types include natural noise and man-made noise. Natural noise includes electrical noise and atmospheric noise. Natural noise originates from noise generated in the natural environment. Man-made noise is generated based on human activities. Man-made noise includes industrial noise, mechanical noise and electronic product noise. The noise in the natural environment and the noise signal categories generated by humans are obtained through big data to obtain corresponding noise signals. By comparing them with the noise signals obtained from big data, the type of current noise can be determined.

6. The method for signal noise suppression and signal integrity protection based on deep learning according to claim 1, characterized in that: The model training method is based on the gradient descent method to update the weights and bias terms, and optimizes the model based on the regularization method to prevent overfitting. The Adam algorithm is selected as the adaptive learning mechanism, the algorithm parameters are configured in the model, and iterative optimization is continuously performed to improve the performance of the neural network model.

7. The method for signal noise suppression and signal integrity protection based on deep learning according to claim 1, characterized in that: When the currently classified noise signal is input, the neural network model is provided with a dual input framework, and different corresponding noise signals are input into the corresponding input entrances of the neural network model, wherein the dual input framework of the neural network model includes a natural noise input entrance and an artificial noise input entrance, and the corresponding input is input into different input entrances after judging the noise of the current input.

8. The method for signal noise suppression and signal integrity protection based on deep learning according to claim 1, characterized in that: When training the model, the parameters and structure of the model are dynamically adjusted according to the performance of the model on the validation set. When training the model, two sets of training sets are prepared, one is an artificial noise data set and the other is a natural noise data set. Each data set is preprocessed. The model contains two parallel sub-networks, one for processing artificial noise data and the other for processing natural noise data.

9. The method for signal noise suppression and signal integrity protection based on deep learning according to claim 6, characterized in that: The iterative training formula is: Among them, θ1, θ2 and θ n is the parameter of the model, L(θ) is the loss function, and the parameters are updated. The update formula is: Where η is the learning rate. In each iteration of the model, the step parameters are updated according to the formula to optimize the performance of the neural network model.

10. The method for signal noise suppression and signal integrity protection based on deep learning according to claim 1, characterized in that: The extracted noise signal is input into the corresponding input port of the trained model. The model suppresses the noise based on the learned knowledge. While suppressing the noise, the model ensures that the integrity of the signal is not affected and retains the useful information in the signal, such as frequency components and phase information.

Images