CN115618215B - Complex electromagnetic environment grading method based on morphological intelligent computation - Google Patents
Complex electromagnetic environment grading method based on morphological intelligent computation Download PDFInfo
- Publication number
- CN115618215B CN115618215B CN202211275076.6A CN202211275076A CN115618215B CN 115618215 B CN115618215 B CN 115618215B CN 202211275076 A CN202211275076 A CN 202211275076A CN 115618215 B CN115618215 B CN 115618215B
- Authority
- CN
- China
- Prior art keywords
- electromagnetic environment
- morphology
- morphological
- signals
- gradient
- 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
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R29/00—Arrangements for measuring or indicating electric quantities not covered by groups G01R19/00 - G01R27/00
- G01R29/08—Measuring electromagnetic field characteristics
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Electromagnetism (AREA)
- Complex Calculations (AREA)
Abstract
The invention relates to a signal processing technology, in particular to a complex electromagnetic environment grading method based on morphological intelligent computation, which adopts a nonlinear method to process electromagnetic environment signals so as to realize electromagnetic environment signal characteristic extraction; the complex electromagnetic environment signals are classified, the electromagnetic environment signals with different complexity are accurately identified, important basis is provided for post-processing such as signal analysis and interference, the electromagnetic environment signals with different complexity are accurately and efficiently classified, and a calculation model and a training algorithm are established; heuristic algorithms are used to initialize training parameters. Extracting morphological characteristics of electromagnetic environment signals by adopting a characteristic extraction method based on logarithmic morphology gradient spectrum; the electromagnetic environment signal characteristic parameters with different complexity extracted by the method have good differentiation. The accuracy of classifying the electromagnetic environment signals is high by adopting a calculation model and a training algorithm established by the dendritic morphology neurons.
Description
Technical Field
The invention relates to a signal processing technology, in particular to a complex electromagnetic environment grading method based on morphological intelligent computation.
Background
With the rapid development of wireless communication technology, the system and modulation patterns of communication signals are complex and various, the frequency spectrums are increasingly crowded and overlapped, the background noise and interference are obviously improved, and the electromagnetic environment is extremely complex.
Such complex electromagnetic environments create serious electromagnetic noise interference and even communication interruption for wireless communication systems, both in the military and civil fields, and thus pose higher requirements and more serious challenges for wireless communication systems, and in particular for signal detection and estimation at the receiving end.
The complex electromagnetic environment is an electromagnetic environment formed by electromagnetic signals with various numbers, complex patterns, dense overlapping and dynamic overlapping distributed on the space domain, the time domain, the frequency domain and the energy in a certain space. The complex electromagnetic environment signal shows the characteristics of typical nonlinearity, non-stationary and strong noise interference. At present, when the feature extraction of electromagnetic environment signals is faced, a plurality of difficulties still exist, so that the classification precision of the electromagnetic signals is not high.
Disclosure of Invention
The invention aims to provide a complex electromagnetic environment grading method based on morphological intelligent computation, which aims to solve the problems in the background technology.
In order to achieve the above purpose, the present invention provides the following technical solutions:
a complex electromagnetic environment grading method based on morphological intelligent computation comprises the following steps:
processing an electromagnetic environment signal by adopting a nonlinear method to realize electromagnetic environment signal characteristic extraction;
step two, classifying complex electromagnetic environment signals, accurately identifying the electromagnetic environment signals with different complexity, and providing important basis for post-processing such as signal analysis, interference and the like; the electromagnetic environment signals with different complexity are accurately and efficiently classified, and a calculation model and a training algorithm are established;
and thirdly, initializing training parameters by using a heuristic algorithm.
The complex electromagnetic environment grading method based on morphological intelligent computation comprises the following steps: in the first step, the mathematical morphology spectrum algorithm adopted for electromagnetic environment signal characteristic extraction is a logarithmic morphology gradient spectrum algorithm.
The complex electromagnetic environment grading method based on morphological intelligent computation comprises the following steps: when the electromagnetic environment signal features are extracted by adopting a logarithmic morphology gradient spectrum algorithm, firstly, proper maximum analysis scale parameters and structural elements are selected, then, multi-scale expansion and corrosion operation are carried out on electromagnetic environment signals with different complexity, the logarithmic morphology gradient spectrum values of the signals are calculated and used as feature vectors of the electromagnetic environment signals, and a feature set of the electromagnetic environment signals based on self-adaptive multi-scale morphology gradient changes and non-negative matrix decomposition, a feature set based on the logarithmic morphology gradient spectrums and a data set formed by the two feature sets are obtained.
The complex electromagnetic environment grading method based on morphological intelligent computation comprises the following steps: in the second step, the electromagnetic environment signals with different complexity are accurately identified by selecting effective features capable of representing feature differences among the electromagnetic environment signals with different complexity.
The complex electromagnetic environment grading method based on morphological intelligent computation comprises the following steps: a calculation model and a training algorithm of a dendritic morphology neural network based on random gradient descent are adopted, and heuristic algorithms are used for initializing learning parameters.
The complex electromagnetic environment grading method based on morphological intelligent computation comprises the following steps: the electromagnetic environment signals are respectively classified based on self-adaptive multi-scale morphological gradient change and non-negative matrix factorization feature sets, feature sets based on logarithmic morphology gradient spectrums and data sets formed by the self-adaptive multi-scale morphological gradient change and non-negative matrix factorization feature sets through a calculation model and a training algorithm of a dendritic morphology neural network based on random gradient descent.
The complex electromagnetic environment grading method based on morphological intelligent computation comprises the following steps: in the random gradient descent based dendritic morphology neural network, the dendritic morphology neurons use superboxes as their classification decision boundaries, one superbox for each dendrite, with the most active dendrite closest to the input class.
The complex electromagnetic environment grading method based on morphological intelligent computation comprises the following steps: a softmax layer is added at the tail end of the dendritic morphology neuron, and the dendritic output is normalized and used as a measurement parameter of class possibility.
The complex electromagnetic environment grading method based on morphological intelligent computation comprises the following steps: in the third step, a heuristic method is adopted to initialize the superbox.
The complex electromagnetic environment grading method based on morphological intelligent computation comprises the following steps: the initialization of the super box adopts HpC initialization and dHpC initialization.
Compared with the prior art, the invention has the beneficial effects that: extracting morphological characteristics of electromagnetic environment signals by adopting a characteristic extraction method based on logarithmic morphology gradient spectrum; the electromagnetic environment signal characteristic parameters with different complexity extracted by the method have good differentiation.
The accuracy of classifying the electromagnetic environment signals is high by adopting a calculation model and a training algorithm established by the dendritic morphology neurons.
Drawings
FIG. 1 is a waveform of a simple electromagnetic environment signal in the present invention.
Fig. 2 is a waveform of a moderately complex electromagnetic environment signal according to the present invention.
FIG. 3 is a waveform of a moderately complex electromagnetic environment signal according to the present invention.
Fig. 4 is a waveform of a moderately complex electromagnetic environment signal of the present invention.
Fig. 5 is a graph of basic mathematical morphology of electromagnetic environment signals of four different complexities of the linear structural elements of the present invention.
FIG. 6 is a graph of basic mathematical morphology of electromagnetic environment signals of four different complexities for square structural elements of the present invention.
FIG. 7 is a graph of basic mathematical morphology of electromagnetic environment signals of four different complexities for diamond-shaped structural elements of the present invention.
Fig. 8 is a morphological gradient spectrum of electromagnetic environment signals of four different complexities of the linear structural elements of the present invention.
FIG. 9 is a morphological gradient spectrum of four electromagnetic environment signals of different complexity for square structural elements of the present invention.
FIG. 10 is a morphological gradient spectrum of four different complexity electromagnetic environment signals of diamond shaped structural elements of the present invention.
FIG. 11 is a log-morphology gradient spectrum of four electromagnetic environment signals of different complexity for the linear structural elements of the present invention.
FIG. 12 is a log-morphology gradient spectrum of four electromagnetic environment signals of different complexity for square structural elements in the present invention.
FIG. 13 is a log-morphology gradient spectrum of four different complexity electromagnetic environment signals for diamond-shaped structural elements of the present invention.
Fig. 14 is a schematic representation of a neuron according to the present invention.
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.
As an embodiment of the present invention, the complex electromagnetic environment grading method based on morphological intelligent computation includes:
and processing the electromagnetic environment signal by adopting a nonlinear method to realize the characteristic extraction of the electromagnetic environment signal.
The complex electromagnetic environment signal shows typical characteristics of nonlinearity, non-stability and strong noise interference; therefore, the processing of the electromagnetic environment signal by adopting a nonlinear method to realize the characteristic extraction of the electromagnetic environment signal is the core and key of the complexity evaluation of the environment signal;
in the invention, the adopted nonlinear method is mathematical morphology, which is a nonlinear analysis method, and the morphological characteristics of signals are depicted from the perspective of collection, so that the logic is strict and the algorithm is simple.
The mathematical morphological spectrum algorithm derived from mathematical morphology is widely applied to the fields of fault diagnosis, image processing and the like by utilizing the mathematical morphological spectrum algorithm to extract nonlinear characteristics, but the application of extracting the characteristics of electromagnetic environment signals is further required to be developed.
The mathematical morphological spectrum algorithm can be used for analyzing the characteristics and structural shapes of images and signals; specifically, structural elements with different scales are adopted to process signals, and structural morphological characteristics of the signals on different scales are reflected.
The mathematical morphology spectrum algorithm comprises a basic mathematical morphology spectrum algorithm, a morphology gradient spectrum algorithm and a logarithmic morphology gradient spectrum algorithm;
and respectively extracting the characteristics of the electromagnetic environment signals by using a basic mathematical morphology spectrum algorithm, a morphology gradient spectrum algorithm and a logarithmic morphology gradient spectrum algorithm, and comparing the results. The result comparison shows that the feature extraction method based on the logarithmic morphology gradient spectrum algorithm is more comprehensive in morphological feature extraction of electromagnetic environment signals, and the extracted feature parameters are better in characterization effect; therefore, the mathematical morphology spectrum algorithm adopted in the invention is preferably a logarithmic morphology gradient spectrum algorithm.
For ease of understanding, the basic mathematical morphology spectral algorithm, morphology gradient spectral algorithm, and logarithmic morphology gradient spectral algorithm are explained below, respectively:
(1) A basic mathematical morphology spectrum algorithm;
the basic mathematical morphology spectrum algorithm is mathematical morphology particle analysis, wherein the mathematical morphology particle analysis is a multi-scale analysis method based on mathematical morphology, and is widely applied to the fields of image processing and mechanical signal processing;
the mathematical morphology particle analysis specifically uses structural elements with different scales and different shapes to filter a research object so as to obtain internal information of the research object;
the mathematical morphology particle analysis is defined by the formula: y= { ψ λ } λ≥0 I.e. a series of psi λ A collection of transforms; transformation psi λ Is defined as:
in the above-mentioned method, the step of,is a morphological open operation symbol, g is a structural element; lambda is the scale parameter of the structural element; thus λg represents a structural element under a certain scale parameter λ; in an open operation environment, the value of lambda is not less than 0;
from the formula, it can be seen that the mathematical morphology particle analysis is based on mathematical morphology open operations.
Wherein, the expression of λg is as follows:
in the above-mentioned method, the step of,is a morphological dilation operation symbol.
The basic mathematical morphology spectrum algorithm takes mathematical morphology particle analysis as an operation basis, the scale parameter lambda is a control variable, and the distribution condition of morphological particles under different scales is intuitively reflected in a curve form.
If a continuous time domain function is defined as f (n) and a structural function as g (m), the basic mathematical morphology spectral algorithm function MS (f, λ, g) is expressed as follows:
in the above formula, s= ≡f (x) dx represents a measure of f (x) in the definition domain; the gray value signal usually takes the ghost area as a measure; therefore, when λ is greater than or equal to 0, the basic mathematical morphology spectrum algorithm corresponding to the function f (n) under the action of the scale structure function λg is also called an open operation mathematical morphology spectrum algorithm, and can be denoted as MS + ；
The mathematical morphology spectrum algorithm corresponding to the open operation mathematical morphology spectrum algorithm is a closed operation mathematical morphology spectrum algorithm and can be marked as MS-; in the closed arithmetic mathematical morphology spectrum algorithm, lambda is less than 0; the expression of the basic mathematical morphology spectral algorithm function MS (f, λ, g) is as follows:
in the above formula, ".
In the morphological particle analysis of one-dimensional discrete signals, the basic mathematical morphological spectral algorithm shows stronger sensitivity in consideration of the numerical variation along with lambda, so lambda can be taken as a continuous integer value, and the maximum value taken by lambda is lambda max And minimum values of lambda respectively min ；
At this time, the mathematical morphology spectrum algorithm of the open operation and the close operation of the one-dimensional discrete signal can be respectively simplified, and the simplified expression is as follows:
MS - (λ,g)＝S'[f·(-λ)g-f·(-λ+1)g] λ min ≤λ＜0；
in the above equation, S' = Σf (n).
Since the open operation mathematical morphology spectrum algorithm has non-expansibility, namelyThe one-dimensional discrete signal's open-computing mathematical morphology spectral algorithm must not be negative, i.e., MS + (lambda, g) is not less than 0; similarly, the closed-form mathematical morphology spectrum algorithm has expansibility, namely f.lambda.1 g is less than or equal to f.lambda.g, namely MS _ (lambda, g) is more than or equal to 0, which ensures that the morphology spectrograms of the one-dimensional discrete signals are all meaningful non-negative spectral lines. Similar to the meaning of the spectrum in fourier transform, the basic mathematical morphology spectrum reflects the distribution of signal morphology features at different structural element scales. For a structural element lambdag under a certain scale, when the signal is more than the corresponding morphological structural component, the spectral line value in the mathematical morphological spectrum of the signal is larger, and conversely, the spectral line value is smaller when the structural component is less than the corresponding morphological structural component.
The basic mathematical morphology spectrum can be simplified into an open operation mathematical morphology spectrum distributed when the scale lambda is more than or equal to 0 and a closed operation mathematical morphology spectrum distributed when the scale lambda is less than 0, wherein the open operation mathematical morphology spectrum reflects the structural characteristic information of the signal, and the closed operation mathematical morphology spectrum reflects the corresponding background information. The duality of the open and close operations can be known: the open-close operation mathematical morphology spectrum is basically consistent in terms of the morphology complexity of the description signal, and in general, the signal structure characteristic information can be fully embodied by only researching the open-operation mathematical morphology spectrum.
(2) A morphological gradient spectrum algorithm;
after morphological dilation and morphological erosion of the time domain f (n) using the structural function g (m), the resulting difference constitutes the concept of a morphological gradient. When the signal is processed, the morphological gradient combines the characteristics of morphological expansion and morphological corrosion, can consider the positive pulse information and the negative pulse information of the signal at the same time, and can effectively extract the morphological characteristics of the signal.
In the morphological gradient spectrum algorithm, the expression of the morphological gradient operator is as follows:
introducing a morphological gradient operator into a basic mathematical morphological spectrum algorithm function to obtain the morphological gradient spectrum algorithm function, wherein the expression is as follows:
for one-dimensional discrete signals, as the morphological gradient operator has expansibility, the morphological gradient spectrum algorithm function can be simplified, and the simplified function expression is as follows:
in the above, MGS + (f, lambda, g) and MGS - (f, lambda, g) are an expanded morphology gradient spectral algorithm function and a corrosion morphology gradient spectral algorithm function, respectively; both reflect the shape change rule of the signal under the scales of different structural elements in the positive and negative intervals respectively. From the duality of the expansion and corrosion operations, it can be seen that the two morphology gradient spectra are essentially identical structures when describing the morphological complexity of the object.
(3) A log morphology gradient spectrum algorithm;
although both the basic mathematical morphology spectrum algorithm and the morphology gradient spectrum algorithm describe the change rule of the shape of the signal on different scales, the basic mathematical morphology spectrum algorithm has statistical deviation, and the morphology gradient spectrum algorithm mainly highlights the pulse information of the signal. For electromagnetic environment signals, the basic mathematical morphological spectrum characteristics of the electromagnetic environment signals under different scales are extracted to find out that the discrimination between the electromagnetic environment signal spectral lines with different complexity is very poor; the morphology gradient spectrum of the electromagnetic environment signals is calculated under different scales, and the obtained curve has a certain distinguishing capability on electromagnetic environment signals with different complexity, but the distinguishing effect is not good.
In order to solve the problems, a logarithmic morphology gradient spectrum algorithm is introduced in the invention, and complex electromagnetic environment signal characteristics based on the logarithmic morphology gradient spectrum are extracted by carrying out logarithmic processing on the morphology gradient spectrum.
Because the expansion morphology gradient spectrum algorithm function MGS is described on the complexity of the signal morphology + (f, lambda, g) and corrosion morphology gradient spectral algorithm function MGS - (f, lambda, g) have consistency, so that the structural characteristic information of the signals is reflected, and therefore only the expansion form gradient spectrum of the signals needs to be considered.
By means of gradient spectrum algorithm function MGS of expansion morphology + And (f, lambda, g) performing logarithmic processing to obtain an expression of the logarithmic morphology gradient spectrum algorithm function on a positive interval, wherein the expression is specifically as follows:
LMGS + ＝log(MGS + (f,λ,g)+1) λ≥0；
LMGS in the above + Log morphology gradient spectral algorithm function.
According to the logarithmic morphology gradient spectrum algorithm function on the positive interval described above, complex electromagnetic environment signal characteristics can be extracted; specifically, firstly, proper maximum analysis scale parameters and structural elements are selected, then multi-scale expansion and corrosion operation are carried out on electromagnetic environment signals with different complexity, logarithmic morphology gradient spectrum values of the signals are calculated according to the above formula and are used as feature vectors of the electromagnetic environment signals, and a feature set of the electromagnetic environment signals based on self-adaptive multi-scale morphology gradient change and non-negative matrix factorization, a feature set based on the logarithmic morphology gradient spectrum and a data set composed of the two are obtained.
The key point of extracting electromagnetic environment signal characteristics by using a mathematical morphology spectrum algorithm is that whether the obtained signal characteristic parameters can reflect signal structure characteristic information comprehensively and simultaneously can effectively distinguish electromagnetic environment signals with different complexity;
therefore, four electromagnetic environment signals with different complexity are simulated for analysis; the sampling frequency is selected to be f=20000 MHz, the sampling time is 2 mu s, and the rationality and the effectiveness of extracting electromagnetic environment signal characteristics based on a logarithmic morphology gradient spectrum algorithm are further verified and analyzed in a Matlab environment.
The time domain waveforms of four electromagnetic environment signals with different complexity are shown in fig. 1-4.
The basic mathematical morphology spectrum MS, the morphology gradient spectrum MGS and the logarithmic morphology gradient spectrum LMGS of four electromagnetic environment signals with different complexity are calculated by using a basic mathematical morphology spectrum algorithm, a morphology gradient spectrum algorithm and a logarithmic morphology gradient spectrum algorithm respectively, and three structural elements of a line shape, a square shape and a diamond shape are selected in the calculation process by comparison analysis.
Basic mathematical morphology spectrograms of electromagnetic environment signals with four different complexities of three structural elements of line shape, square shape and diamond shape can be seen in fig. 5-7; the morphological gradient spectrogram can be seen in fig. 8-10; the log morphology gradient spectrum can be seen in fig. 11-13.
By comparing the figures 5 to 7, it can be obtained that different structural elements have similar treatment effects on the obtained basic mathematical morphology spectrum MS curve; it is shown that the influence of linear, square and diamond structural elements on the time domain signal of the electromagnetic environment signal is similar when the basic mathematical morphological spectral feature extraction is performed on the time domain signal.
As can be seen from the curve analysis of fig. 8-13, the effects of the linear, square, and diamond-shaped structural elements on the morphology gradient spectrum and the logarithmic morphology gradient spectrum are similar.
As can be seen from fig. 11, the log-morphology gradient spectrum LMGS curves of the four electromagnetic environment signals with different complexity have good distinguishability, the spectrum value of each curve monotonically decreases with the increase of the scale parameters of the structural elements, and the variation trend is consistent. And as the spectrum value is continuously reduced, the difference value between the spectrum curves of each environmental signal is not obviously reduced, which indicates that the feature extraction algorithm based on the log-morphology gradient spectrum LMGS has good stability.
In fig. 5, although the overall basic mathematical morphology spectrum MS curve shows a decreasing trend, the basic mathematical morphology spectrum MS curve does not decrease steadily with increasing scale parameters, the up-down fluctuation of the spectrum values is obvious, and after the scale parameters are greater than 15, different signal spectrums are already staggered, so that it can be seen that the electromagnetic environment signals with different complexity cannot be distinguished when the morphology spectrum values calculated by the basic mathematical morphology spectrum algorithm are taken as the characteristic values.
In fig. 8 to 10, the change trends of the four morphology gradient spectrum MGS curves are uniform, and all show monotonous decreasing trends, and there is a certain degree of discrimination when the scale is small, but the degree of discrimination of the four curves decreases with increasing scale.
Therefore, the electromagnetic environment signal characteristic parameters with different complexity extracted by the method have good distinguishing performance.
In the method, complex electromagnetic environment signals are classified, electromagnetic environment signals with different complexity are accurately identified, and important basis is provided for post-processing such as signal analysis, interference and the like; and accurately and efficiently classifying electromagnetic environment signals with different complexity.
The electromagnetic environment signals with different complexity are accurately identified by selecting effective features capable of representing feature differences among the electromagnetic environment signals with different complexity.
Extracting the characteristics of electromagnetic environment signals with different complexity through a logarithmic morphology gradient spectrum algorithm;
accurately identifying the extracted features of the electromagnetic environment signals with different complexity, and selecting and designing a classifier to intelligently classify the accurately identified features.
For this purpose, the invention applies the dendritic morphology neural network to the recognition of electromagnetic environment signals.
Wherein the dendritic morphology neural network comprises a plurality of dendritic morphology neurons, the dendritic morphology neurons being an artificial neural network using minimum and maximum operators instead of algebraic products; in dendritic morphology neurons, morphology operators build superboxes in N-dimensional space; these superboxes build heuristic-based training methods without using optimization methods;
because the heuristic training method is adopted, the convergence problem does not exist, the perfect classification can be achieved, and the training speed is very high. However, perfect classification is inconvenient in practical application, because it can lead to excessive fitting of training data, increasing complexity of learning models;
in order to solve the problems, the invention adopts a calculation model and a training algorithm of a dendritic morphology neural network based on random gradient descent, applies the model to classification of electromagnetic environment signals with different complexity, and only uses the heuristic algorithms to initiate learning parameters.
In the traditional artificial neural network, the output of the neuron is obtained by performing addition, subtraction, multiplication and division on the input of the neuron;
in a conventional artificial neural network, the calculation formula of neurons is as follows:
in the above, τ j Representing the output of the jth neuron, x i A value representing an ith neuron connected to the jth neuron, which is a set of real numbers; w (w) ij Representing the connection weights of two neurons, θ j Representing the activation threshold of the jth neuron.
The schematic diagram of the neuron is shown in fig. 14.
Unlike conventional artificial neural networks, dendritic morphology neurons are nonlinear neurons, which are not linear algorithms such as addition, subtraction, multiplication, division and the like in conventional artificial neural networks, but lattice algebraic systems instead, so that the combination of mathematical morphology and conventional artificial neural networks is realized.
Dendritic morphological neurons have distinct post-dendritic regions that accept axonal terminal branch inputs from other neurons, whereas the post-synaptic membrane of dendrites adopts a stimulus or inhibition response to the input signals it receives. Neurons react to the total dendritic input and are activated by a function.
Dendritic morphology neurons use a superbox as their classification decision boundary, one for each dendrite, with the most active dendrite closest to the input class.
In the classical perceptron domain, the most important training method is gradient descent, and has been applied to training multi-layer perceptrons, developing back propagation algorithms; while most morphological neuron training methods are based on intuition rather than gradient descent;
although the existing training method of morphological neurons is also based on gradient descent, the training method is only suitable for regression tasks.
The expansion of the above method in classification is not straightforward due to morphological non-variability;
for this reason, a softmax layer is added at the tail end of the dendritic morphology neuron in the invention; thereby deriving a dendritic morphology neuron based on random gradient descent training; the softmax layer is a flexible maximum transfer function layer and is a category of neural network classification layers.
In the invention, the dendritic outputs are normalized by adding a softmax layer at the tail end of the dendritic morphological neuron, so that the outputs are limited to be between 0 and 1; these outputs can be used as a measure of class likelihood so that it can be determined which of the classes the input vector belongs to.
In conventional dendritic morphology neurons, one superbox is represented by its extreme points; and in the present invention is represented by the minimum pole and its size vector.
In the training of the morphological neural network, the number of the superboxes and the dendritic parameters are automatically determined. This is in contrast to conventional sensors, where only the learning parameters are determined during the training process.
In contrast, the invention initializes the superbox based on a heuristic algorithm and optimizes the dendritic parameters using random gradient descent based.
For the initialization of the superbox using the heuristic-based method, hpC initialization and dHpC initialization methods are used.
In the HpC initialization method, each training set of the same class is contained in a superbox; the maximum number of dendrites is equal to the number of classes.
The dHpC initialization method is used to partition each superbox generated by HpC.
In the invention, a logarithmic morphology gradient spectrum algorithm is adopted to perform feature extraction on an electromagnetic environment signal sample, so as to obtain a feature set of the electromagnetic environment signal based on self-adaptive multi-scale morphology gradient change and non-negative matrix factorization, a feature set based on logarithmic morphology gradient spectrum and a data set formed by the two;
then, classifying the three characteristic parameter sets by using dendritic morphology neurons based on random gradient descent training;
and finally, initializing the dendritic morphology neurons based on random gradient descent training.
The specific classification results are as follows:
500 samples were taken for each complex electromagnetic environment signal, for a total of 2000 samples. When the DMN-SGD is used for carrying out complexity identification and classification on electromagnetic environment signals, 400 sample features are randomly selected from the electromagnetic environment signals with each complexity to serve as training samples, and 100 samples are left to serve as test samples, so that accurate and stable experimental results are obtained, and the obtained experimental data are average values of 20 classification results.
The DMN-SGD is initialized by adopting K-means, hpC, dHpC respectively and is analyzed and compared with a support vector machine SVM and a multi-layer perceptron MLP.
TABLE 1 electromagnetic environmental Signal Classification results
Table 1 shows classification results of 5 kinds of classifiers. From the data in the table, it can be seen that the DMN-SGD model proposed herein can achieve very high classification accuracy when classifying the complexity of the electromagnetic environment signal. The classification results will also vary correspondingly using different initialization methods. But the classification accuracy of the DMN-SGD is higher than that of the SVM and the MLP in the whole.
The above-described embodiments are illustrative, not restrictive, and the technical solutions that can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention are included in the present invention.
Claims (5)
1. The complex electromagnetic environment grading method based on morphological intelligent computation is characterized by comprising the following steps of:
processing an electromagnetic environment signal by adopting a nonlinear method to realize electromagnetic environment signal characteristic extraction;
step two, classifying the complex electromagnetic environment signals, and accurately identifying the electromagnetic environment signals with different complexity; classifying electromagnetic environment signals with different complexity, and establishing a calculation model and a training algorithm;
the electromagnetic environment signals with different complexity are accurately identified by selecting effective features capable of representing feature differences among the electromagnetic environment signals with different complexity;
adopting a calculation model and a training algorithm of a dendritic morphology neural network based on random gradient descent, and using a heuristic algorithm to initially learn parameters;
the dendritic morphology neural network comprises a plurality of dendritic morphology neurons, wherein morphology operators construct superboxes in an N-dimensional space, the dendritic morphology neurons use the superboxes as classification decision boundaries, each dendrite can generate one superbox, and the number of the superboxes and dendrite parameters are automatically determined;
for initializing the super box by adopting a heuristic method, adopting HpC initialization and dHpC initialization methods; in the HpC initialization method, each training set of the same class is contained in a superbox; the maximum number of dendrites is equal to the number of classes, and the dHpC initialization method is used to partition each superbox generated by HpC;
adding a softmax layer at the tail end of the dendritic morphology neuron, normalizing the dendritic output and taking the normalized dendritic output as a measurement parameter of class possibility;
the dendritic morphological neurons have different post-synaptic regions, and the post-synaptic membrane of the dendrites adopts a stimulus or inhibition response to an input signal received by the post-synaptic membrane;
and thirdly, initializing training parameters by using a heuristic algorithm.
2. The method for classifying complex electromagnetic environment based on morphological intelligent computation according to claim 1, wherein in the first step, the mathematical morphological spectral algorithm adopted for electromagnetic environment signal feature extraction is a logarithmic morphological gradient spectral algorithm.
3. The complex electromagnetic environment grading method based on morphological intelligent computation according to claim 2, wherein when the electromagnetic environment signal features are extracted by adopting a logarithmic morphology gradient spectrum algorithm, firstly, the largest analysis scale parameters and structural elements are selected, then, multi-scale expansion and corrosion operation are carried out on electromagnetic environment signals with different complexity, and logarithmic morphology gradient spectrum values of the signals are calculated and used as feature vectors of the electromagnetic environment signals, so that a data set which is formed by the electromagnetic environment signals based on self-adaptive multi-scale morphology gradient change and non-negative matrix factorization feature sets, the feature sets based on the logarithmic morphology gradient spectrums and the feature sets is obtained.
4. The method for classifying the complex electromagnetic environment based on morphological intelligent computation according to claim 2, wherein the electromagnetic environment signals are classified based on adaptive multi-scale morphological gradient change and non-negative matrix factorization feature sets, feature sets based on logarithmic morphology gradient spectrums and data sets composed of the two by a computation model and a training algorithm of a dendritic morphology neural network based on random gradient descent.
5. The method for classifying complex electromagnetic environments based on morphological intelligent computation according to claim 1, wherein in the third step, a heuristic method is adopted to initialize the superbox.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211275076.6A CN115618215B (en) | 2022-10-18 | 2022-10-18 | Complex electromagnetic environment grading method based on morphological intelligent computation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211275076.6A CN115618215B (en) | 2022-10-18 | 2022-10-18 | Complex electromagnetic environment grading method based on morphological intelligent computation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115618215A CN115618215A (en) | 2023-01-17 |
| CN115618215B true CN115618215B (en) | 2023-09-01 |
Family
ID=84863385
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211275076.6A Active CN115618215B (en) | 2022-10-18 | 2022-10-18 | Complex electromagnetic environment grading method based on morphological intelligent computation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115618215B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111738128A (en) * | 2020-06-17 | 2020-10-02 | 山东卓文信息科技有限公司 | Series fault arc detection method based on morphological filtering and MMG |
| CN114217286A (en) * | 2021-12-10 | 2022-03-22 | 中国人民解放军63893部队 | Radar unintentional modulation signal extraction method based on form component |
-
2022
- 2022-10-18 CN CN202211275076.6A patent/CN115618215B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111738128A (en) * | 2020-06-17 | 2020-10-02 | 山东卓文信息科技有限公司 | Series fault arc detection method based on morphological filtering and MMG |
| CN114217286A (en) * | 2021-12-10 | 2022-03-22 | 中国人民解放军63893部队 | Radar unintentional modulation signal extraction method based on form component |
Non-Patent Citations (1)
| Title |
|---|
| Photovoltaic Power Station Electromagnetic Environment Complexity Evaluation Utilizing Logarithmic Morphological Gradient Spectrum;Hua-Chen Xi 等;《Frontiers in Energy Research》;正文第1-8页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115618215A (en) | 2023-01-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Patil et al. | White blood cells image classification using deep learning with canonical correlation analysis | |
| CN110728360B (en) | Micro-energy device energy identification method based on BP neural network | |
| Ghosh et al. | A neural network based hybrid system for detection, characterization, and classification of short-duration oceanic signals | |
| CN109919241B (en) | Hyperspectral unknown class target detection method based on probability model and deep learning | |
| CN107844751B (en) | Method for classifying hyperspectral remote sensing images of guide filtering long and short memory neural network | |
| Matuszewski et al. | Neural network application for emitter identification | |
| CN112001270A (en) | Ground radar automatic target classification and identification method based on one-dimensional convolutional neural network | |
| CN109766791B (en) | Communication signal modulation identification method based on self-encoder | |
| CN110647788B (en) | Human daily behavior classification method based on micro-Doppler characteristics | |
| Li et al. | Automatic modulation classification based on bispectrum and CNN | |
| Bodyanskiy | Computational intelligence techniques for data analysis | |
| Yang et al. | Open set recognition of underwater acoustic targets based on GRU-CAE collaborative deep learning network | |
| Shen et al. | Node identification in wireless network based on convolutional neural network | |
| CN114964778A (en) | Bearing fault diagnosis method based on wavelet time-frequency graph and deep learning | |
| Yu et al. | Fingerprint extraction and classification of wireless channels based on deep convolutional neural networks | |
| Kirichenko et al. | Two approaches to machine learning classification of time series based on recurrence plots | |
| CN113052099B (en) | SSVEP classification method based on convolutional neural network | |
| Hu et al. | An features extraction and recognition method for underwater acoustic target based on ATCNN | |
| Zhang et al. | Target detection for HFSWR based on an S 3 D algorithm | |
| Chinmayi et al. | Emotion Classification Using Deep Learning | |
| CN117606801A (en) | Cross-domain bearing fault diagnosis method based on multi-characterization self-adaptive network | |
| CN115618215B (en) | Complex electromagnetic environment grading method based on morphological intelligent computation | |
| Xiao et al. | Health assessment for piston pump using LSTM neural network | |
| CN114692773A (en) | End-to-end deep learning Raman spectrum data classification method based on DRS-VGG | |
| Aslinezhad et al. | A noise-aware deep learning model for automatic modulation recognition in radar signals |
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 | ||
| GR01 | Patent grant |