CN110632386B - Solar radio interference filtering method, readable storage medium and electronic equipment - Google Patents
Solar radio interference filtering method, readable storage medium and electronic equipment Download PDFInfo
- Publication number
- CN110632386B CN110632386B CN201910758704.8A CN201910758704A CN110632386B CN 110632386 B CN110632386 B CN 110632386B CN 201910758704 A CN201910758704 A CN 201910758704A CN 110632386 B CN110632386 B CN 110632386B
- Authority
- CN
- China
- Prior art keywords
- radio
- solar
- data
- signal
- radio station
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000001914 filtration Methods 0.000 title claims abstract description 41
- 238000013528 artificial neural network Methods 0.000 claims abstract description 32
- 238000012549 training Methods 0.000 claims abstract description 17
- 125000004122 cyclic group Chemical group 0.000 claims abstract description 14
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 230000000306 recurrent effect Effects 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 11
- 230000015654 memory Effects 0.000 claims description 11
- 238000004880 explosion Methods 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 7
- 230000008034 disappearance Effects 0.000 claims description 5
- 230000007787 long-term memory Effects 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 230000006403 short-term memory Effects 0.000 claims description 4
- 238000009825 accumulation Methods 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 claims description 3
- 230000015572 biosynthetic process Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000010287 polarization Effects 0.000 claims description 3
- 238000003786 synthesis reaction Methods 0.000 claims description 3
- 108010076504 Protein Sorting Signals Proteins 0.000 claims description 2
- 238000012216 screening Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 7
- 238000012545 processing Methods 0.000 description 15
- 230000008569 process Effects 0.000 description 7
- 238000013135 deep learning Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 3
- 238000003672 processing method Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000005855 radiation Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 238000011425 standardization method Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R23/00—Arrangements for measuring frequencies; Arrangements for analysing frequency spectra
- G01R23/16—Spectrum analysis; Fourier analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Monitoring And Testing Of Transmission In General (AREA)
Abstract
The utility model provides a solar radio interference filtering method, a readable storage medium and an electronic device, which collects the intensity data of the set time period before and after the solar burst event, and screens out the frequency channel where the radio station signal meeting the conditions is located through the radio flow value data of all frequency channels at a certain moment in the non-burst period; preprocessing the collected radio station signal data meeting the conditions to establish a data set; establishing a cyclic neural network, training the cyclic neural network by utilizing the preprocessed radio station signal data, and predicting a signal value of a radio station in a solar burst area by using the trained cyclic neural network; according to the linear additivity of the signals, subtracting the predicted radio station value from the original value of the radio current at the corresponding moment of the burst area to obtain the value of a clean pure solar burst event; the method can predict the radio station interference signals more accurately, further retain more solar burst information, and the final solar burst observation effect is obviously superior to that of a common image filtering method.
Description
Technical Field
The disclosure relates to the technical field of solar radio frequency spectrometers, in particular to a solar radio interference filtering method, a readable storage medium and a solar radio frequency spectrometer.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
The space observation of the solar activity needs to use a solar radio frequency spectrometer, but the solar radio frequency spectrometer can receive various radio station interference signals and other electromagnetic interference signals existing in the space while receiving solar radio signals, the existence of the interference signals can not obtain a clean and clear solar radio dynamic spectrogram, especially, some interference signals which occupy wider frequency bands and have signal intensity greater than the solar radio flow intensity can cover an explosion event, and therefore the observation and analysis of the solar radio explosion event are seriously influenced.
In order to better observe the complete solar burst event, the interference filtering measures are often divided into hardware processing and software processing. For example, interference can be suppressed by selecting an anti-interference amplifier, a filter and the like in hardware processing, but the hardware processing method is not suitable for the established solar radio frequency spectrometer because the circuit element adjustment period is long, the processing efficiency is low, and various interference signals cannot be removed completely by using the method. Therefore, in general, each observation station chooses to process the interference signal by using software. For example, wavelet transformation is used for removing interference processing aiming at different burst events, and finally burst structures are extracted, but the method is mostly used for processing a grayscale map of a frequency spectrum, and human factors exist for selection of interference signals and threshold values, so that the method is not suitable for filtering processing of a plurality of burst events; or a calibration method is adopted and the result is given, so that the possibility of obtaining a clear solar radio frequency spectrogram is provided, but the calibration of the spectrometer needs a large amount of historical data and has certain errors. Therefore, no better method for effectively removing the radio station interference signals in the solar burst area exists at present.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a solar radio interference filtering method, a readable storage medium and a solar radio frequency spectrograph, wherein a circulating neural network in deep learning is used for predicting radio station interference signals in a solar burst area, the radio station interference signals in any frequency band and any time period can be trained and predicted, more effective information of a solar burst event can be reserved, a larger operation space is provided for the analysis of a subsequent solar burst event, and a new thought and direction are provided for deep learning in astronomical anti-interference processing.
In order to achieve the purpose, the following technical scheme is adopted in the disclosure:
in a first aspect, the present disclosure provides a solar radio interference filtering method;
a solar radio interference filtering method comprises the following steps:
(1) collecting intensity data of a set time period before and after a solar burst event, screening out frequency channels where radio station signals meeting conditions are located through radio flow rate value data of all frequency channels at a certain time of a non-burst time period, and positioning the position of the initial solar burst time by utilizing a radio flow rate change curve which changes along with time under a single frequency channel;
(2) preprocessing the collected radio station signal data meeting the conditions to establish a data set;
(3) establishing a cyclic neural network, training the cyclic neural network by utilizing the preprocessed radio station signal data, and predicting a signal value of a radio station in a solar burst area by using the trained cyclic neural network;
(4) and according to the linear additivity of the signals, subtracting the predicted station value from the original value of the radio current at the corresponding moment of the burst area to obtain the value of the clean pure solar burst event.
As some possible implementations, in the step (1),
(1-1) drawing radio flow of all frequency points at a certain moment in a non-outbreak period;
(1-2) determining an intensity threshold according to the magnitude of the radio current value of the explosion event;
and (1-3) selecting the frequency channel with the intensity value larger than the intensity threshold value, and storing the intensity value of the selected radio station signal according to the corresponding frequency channel.
As some possible implementation manners, in the step (1), after the radio station signal to be processed is screened out, a certain frequency channel of the radio station signal is selected, and the position of the initial moment of the solar radio burst is located according to a curve of the radio flow of the frequency channel changing with time.
As some possible implementation manners, in the step (2), the acquired data is subjected to FFT operation and digital polarization synthesis operation in the FPGA, and a left-handed signal and a right-handed signal are finally obtained;
reading data according to a set data format, performing accumulation operation in time resolution, and selecting one of a left-handed signal and a right-handed signal as a signal to be processed;
further, a min-max standardization method is adopted to carry out normalization processing on the training data, and the data are linearly transformed to [0, 1 ];
further, the radio station signal sequence is segmented in time to establish a mapping relationship between time segments.
As some possible implementation manners, in the step (3), pure electric station data before the initial moment of solar burst is firstly divided into a training set and a test set, a network trained by the training set is used for predicting station data, the predicted value is compared with the original pure electric station data at the corresponding moment, and the structure of the recurrent neural network is adjusted and optimized according to the comparison result.
As some possible implementation manners, in the step (3), all the radio station data before the initial moment of the solar burst are used as a training set, the radio station data are trained by using the adjusted and optimized recurrent neural network structure, and the radio station signal value of the burst area is predicted.
As some possible implementation manners, in the step (3), several fixed types of values of the station of the single frequency channel are digitally mapped by using a distribution characteristic rule of the station interference signal values, so that the prediction problem is firstly converted into a classification problem to be processed, so as to reduce an error caused by the step-wise prediction.
As some possible implementation manners, in the step (3), the recurrent neural network includes an input layer, a hidden layer and an output layer, and in the forward transmission process of data, the hidden layer receives not only the input layer information at the current moment, but also the information of a hidden layer unit at the previous moment, so as to realize the memory of the information characteristics at the past moment.
As some possible implementation manners, in the step (3), the recurrent neural network is a long-term and short-term memory network, and an information gate for judging information is added in a module repeatedly linked in a hidden layer of the long-term and short-term memory network, so that disappearance of a gradient caused by an excessive number of layers is avoided.
By way of further limitation, in step (3), the information gate is generated by a Sigmoid function, and includes an input gate, a forgetting gate and an output gate, and each information passing through the information gate has a real parameter in the range of [0, 1], where the real parameter represents the proportion of information passing through each gate.
As some possible implementation manners, in the step (4), the signal values of the non-solar burst area radio stations are processed by using a common image filtering method.
In a second aspect, the present disclosure provides a readable storage medium, on which a program is stored, which when executed by a processor, implements the steps of the solar radio interference filtering method described in the present disclosure.
In a third aspect, the present disclosure provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor, where the processor executes the program to implement the steps of the solar radio interference filtering method according to the present disclosure.
As some possible implementations, the electronic devices described in the present disclosure include, but are not limited to, computers, solar radio spectrometers, various microprocessors, processors, servers, or various solar radio observation devices.
Compared with the prior art, the beneficial effect of this disclosure is:
the method utilizes a cyclic neural network in deep learning to predict the radio station interference signals in an outbreak area, and subtracts the predicted values of the radio station signals in the area from the numerical values of the outbreak area containing the radio station interference according to the linear summable principle of the signals, thereby achieving the purpose of removing the interference; the radio station interference signal of any frequency band and any time period can be trained and predicted by utilizing the trained network model, more effective information of solar outbreak events can be reserved, a larger operation space is reserved for subsequent event analysis, and a new thought and direction are provided for deep learning in astronomical anti-interference treatment.
Compared with the traditional prediction method, the prediction method can more accurately predict the radio station interference signal, further more solar burst information is reserved, and the final solar burst observation effect is obviously superior to that of the common image filtering method.
Theoretically, a sufficiently long time sequence can be processed in a circulating mode, but the problem of gradient disappearance enables the RNN to process only short-time information in practice, and the radio station signal disclosed by the disclosure needs to process a long time sequence at one time, so that the radio station signal is predicted by adopting a long-time memory network, and the problem of gradient disappearance caused by too many layers is avoided by adding an information gate for judging information in a module repeatedly linked in a hidden layer.
The method includes the steps that the station signal value of the outbreak area is finally predicted, the original numerical value of the prediction area cannot be obtained in an actual outbreak event, namely the accuracy of a network prediction result cannot be checked, multi-frame station data in front of the prediction area are firstly predicted, the prediction numerical value is compared with original data at a corresponding moment, and a network structure is adjusted and optimized according to the comparison result, so that the prediction result of the recurrent neural network is closer to the real value, and more accurate prediction is achieved.
Drawings
Fig. 1 is a flowchart of a solar radio interference filtering method in embodiment 1 of the present disclosure.
Fig. 2 is a graph of solar radio burst event intensity in example 1 of the present disclosure.
Fig. 3 is a radio flow chart (the time without solar radio burst) at different frequency channels at a single time in embodiment 1 of the present disclosure.
Fig. 4 is a positioning diagram of the initial moment of solar radiation explosion in embodiment 1 of the present disclosure.
Fig. 5 is a schematic diagram of data set creation in embodiment 1 of the present disclosure.
Fig. 6 is a schematic structural diagram of a recurrent neural network in embodiment 1 of the present disclosure.
Fig. 7 is a diagram of an internal structure of a hidden layer of the LSTM recurrent neural network in embodiment 1 of the present disclosure.
Fig. 8(a) and 8(b) are a line graph and a scatter graph of the prediction results of the test set (100 frames of station data) of one frequency channel of the 244MHz station, respectively, in which dotted lines and asterisk scatters indicate the predicted data, and solid lines and dots indicate the original data of the station.
Fig. 9(a), (c), and (e) show the original image of the 240MHz station signal, the normal image filtering result, and the filtering result according to the method described in example 1 of the present disclosure.
Fig. 9(b), (d), and (f) show the original image of the 244MHz station signal in embodiment 1 of the present disclosure, the filtering result of the normal image, and the filtering result of the method described in embodiment 1.
Fig. 10 is a graph of the results of a complete burst event processing in example 1 of the present disclosure.
Fig. 11 is a graph of the results after a complete burst event has reduced resolution in example 1 of the present disclosure.
Detailed Description
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present disclosure. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
Example 1:
as shown in fig. 1, embodiment 1 of the present disclosure provides a method for filtering solar radio interference, and this embodiment uses a recurrent neural network to process an interference signal in solar radio, and mainly includes two stages, namely, predicting a station signal value in an outbreak area and subtracting a predicted value from the outbreak area by the network.
In the first stage, a data set is established, the data set is preprocessed and then enters a built network for training, then the network structure is continuously optimized according to a training result, the network hyper-parameter is adjusted, and finally the test set enters a test network for numerical prediction. In the second stage, the present embodiment filters out the solar radio frequency spectrum interference signal, and the interference signal processed by using the recurrent neural network is a radio station signal occupying a wider frequency band or having a signal intensity close to or even greater than the radio flow rate during solar burst.
(1) Source of station data
Different from common image processing, the method used in the embodiment is performed on a data level, the data selected in the embodiment is a data packet acquired by a racking-meter-band high-resolution solar radio receiver independently developed by a project group, the data in the data packet is subjected to FFT (fast Fourier transform) operation and digital polarization synthesis operation in an FPGA (field programmable gate array), then accumulation operation is performed within time resolution, and a left-handed signal and a right-handed signal are finally obtained. Because the two signals have equal values and opposite directions, only the left-handed signal is selected for research.
The combed mountain observation station observes a solar burst event in 2017, 9 and 9 days, and the quiet solar radiation value of the event is 0, namely the image background value is 0, so that the burst event containing a radio station interference signal in a period of the day is selected for filtering in the embodiment, fig. 2 is a selected burst event intensity graph, the frequency range of the graph is 180MHz-330MHz, the frequency resolution is 128kHz, the time resolution is 10ms, the horizontal coordinate in fig. 2 is time, the vertical coordinate represents frequency, the middle part of the graph represents a solar radio current value and a radio station signal intensity value, different color depths represent different numerical values, and the specific numerical value can refer to the right color axis.
The station interference signals in fig. 2 are mainly concentrated in 240MHz-270MHz, and the station with high signal strength directly covers the burst event, which affects the observation and research of the complete burst event. In this embodiment, the radio signals are selected and processed by using a cyclic neural network method, the radio flow rates of all frequency points at a certain time (time without a burst event) in fig. 2 are firstly plotted, and as a result, as shown in fig. 3, then a threshold is defined reasonably according to the magnitude of the radio flow rate value of the burst event, a frequency channel greater than the threshold is selected (the threshold selected in this embodiment is > -3), and finally, the value of the selected radio signals is stored according to the corresponding frequency channel, so as to establish a data set.
After the station signal to be processed is screened out, the initial moment of the burst is determined. A certain frequency point near a radio station is selected, the position of the solar radio burst is positioned according to the curve of the radio flow of the frequency point along with the time change, the curve of the radio flow value along with the time change of the frequency point is shown in figure 4 by taking 260MHz as an example, the time of the solar burst is international time 6:52:23.755.3, and therefore, the data before the time is taken as training data.
(2) Creation of data sets
The establishment of the data set and the establishment of the network form two main parts of a deep learning algorithm. In order to improve the accuracy of the training and prediction results, preprocessing is required before data enters the network. Although the original data units of the station signals extracted previously are consistent, there exists a span between the data, which causes the convergence speed of the training network to become slow, and the final effect is not ideal. Therefore, this embodiment uses min-max normalization to normalize the training data and linearly transform the data to [0, 1 ].
The prediction of the radio station sequence can be understood as regression analysis, that is, trend prediction is performed by analyzing the relationship between sequence values, but actually, radio station data acquired by a spectrometer may change at a certain moment due to the influence of external factors, so that the prediction of a final result is influenced, and therefore, under the condition that the sample data amount is sufficient, the radio station sequence is segmented, mapping is established, and the relationship between a time period and a time period is found in the embodiment.
FIG. 5 is a schematic of dataset creation, where xtRepresenting the station value at time t. Picture packetThe system comprises an input window, an output window and a sectional sliding window; the segmented sliding window comprises all sequences of an input window and an output window, and the three windows as a whole slide on a time axis to take values; the input window and the output window contain the same number of data frames, which is named time step as one of the network super parameters. In addition, a certain time interval exists between two windows, and in order to keep the relationship between adjacent sequences as much as possible and improve the accuracy of prediction, the time interval is determined to be 1 frame in the embodiment. Therefore, a mapping relation of the segments is established, namely sequence values from (t- (time _ step-2)) to (t +1) are predicted according to the sequence conditions from (t- (time _ step-1)) to (t), wherein the sequence value at the time of (t +1) is a prediction result of the required station sequence.
(3) Establishment of recurrent neural networks
The recurrent neural network introduces the concept of time through circulation, so that data is continuously accumulated on a time axis, and historical information is stored. FIG. 6 is a diagram of a conventional neural network, in which A is a block of the neural network, xt、ht、ytRespectively representing the input, hidden layer state and output at time t, w1、w2、w3Is a matrix of weight coefficients between layers. As can be seen from the figure, in the process of forward transmission of data, the hidden layer unit not only receives the input layer information of the current moment, but also receives the information of the hidden layer unit of the previous moment, so as to realize the memory of the information characteristics of the past moment.
The hidden layer output can be expressed as:
ht＝f(ht-1，xt) (1)
where f (·) is an activation function of the hidden layer, a tanh function and a Sigmoid function are often used, and the latter is used in the present embodiment. The network output may be expressed as:
yt＝g(w3ht) (2)
where g (-) is the activation function of the output layer.
In theory the recurrent neural network can handle long enough time series, but the problem of gradient vanishing makes RNNs in practice only able to handle short periods of information. The station signal described in this embodiment needs to be processed once for a Long time sequence, so a special cyclic neural network structure, Long-Short-Term Memory network (LSTM), is selected to predict the station signal.
The gradient disappearance problem caused by too many layers is avoided by adding a 'gate' for judging information in a module repeatedly linked in the hidden layer. The information gate is generated through a Sigmoid function and is divided into an input gate, a forgetting gate and an output gate. The nature of the function is such that each pass-gate message has a real parameter in the range of 0, 1, which represents the fraction of the message that passes through each gate. FIG. 7 is a block diagram of the inside of the hidden layer of the LSTM at a certain time, illustrating the flow process of data.
The above data transfer process can be expressed by a set of formulas, forgetting the formula:
ft＝σ(Wf·[ht-1，xt]+bf) (3)
input gate calculation formula:
it＝σ(Wi·[ht-1，xt]+bi) (4)
output gate calculation formula:
ot＝σ(Wo·[ht-1，xt]+bo) (7)
ht＝ot*tanh(Ct) (8)
wherein W generally refers to weight parameters of the network, b generally refers to bias, t represents time, h is a hidden layer state, x is input of the network, C refers to LSTM cell state, i, f and o respectively represent input gate, forgetting gate and output gate structures, activation functions of the three gates are Sigmoid functions, and when the network updates the cell state, a tanh function is selected.
The final prediction of the embodiment is the station signal value of the burst area, and since the original value of the prediction area cannot be known in the actual burst event, the accuracy of the network prediction result cannot be checked. Therefore, the chapter firstly predicts the radio station data of 100 frames before the prediction area, compares the prediction value with the original data of the corresponding position, and adjusts and optimizes the network structure according to the comparison result. Fig. 8(a) and 8(b) are a line graph and a scatter plot, respectively, of the predicted results of a test set (100 frames of station data) for one of the frequency channels of a 244MHz station. As can be seen from fig. 8(a), the trained network can predict most of the station signal variation trends, and although the accuracy of the prediction values in fig. 8(b) is not particularly high, the results are not much different from the original values.
Predicting all the radio station signals of the event in the embodiment by using the network structure and the network parameters of the test results of fig. 8(a) and fig. 8(a), then processing by using Matlab software, and subtracting the radio station value predicted by using a recurrent neural network method from the original value of the solar burst area; the non-burst region is filtered using normal image processing methods, i.e. the average of the frequency channel is subtracted. FIGS. 9(a), 9(c) and 9(e) show the original image of the 240MHz station signal, the filtering result of the normal image and the filtering result of the method of the present embodiment; fig. 9(b), 9(d) and 9(f) show the original image of the 244MHz station signal, the normal image filtering result and the filtering result according to the method of the present embodiment. By comparison, the processing method of the embodiment retains more outbreak information, and the effect is obviously better than that of the common image filtering method.
Meanwhile, some narrow-band interference signals with weak signal strength and without great influence on observation of solar burst events can be processed by using a common image filtering method.
The results of the station jammer signal processing in a full solar burst event are shown in fig. 10. It can be seen from fig. 10 that the interference signal processing is cleaner and after the interference is removed, small burst events within the rectangular box are clearly visible. But the non-burst area processed by the common image filtering method has some sporadic radio station interference signals, and the display effect can be weakened by reducing the time and frequency resolution. Fig. 11 is a graph of the effect of reducing the time resolution by 6 times (i.e., down-sampling to 60ms) and the frequency resolution by 16 times (i.e., down-sampling to 0.49MHz), and it can be seen that in the same color scale axis range, sporadic station signals are hardly visible (actually, there is also a value), and a complete burst event is clearly displayed.
In the embodiment, a special recurrent neural network-LSTM network is adopted to predict the radio station interference signal in the solar radio burst area, and the predicted radio station value is subtracted from the value of the solar burst area according to the linear additivity of the signal to achieve the purpose of filtering. The method of the embodiment is based on data processing, can furthest retain effective information related to the outbreak event, and the trained network can predict the radio station value of any frequency band and any time, so that a new idea is provided for deep learning in astronomy anti-interference processing.
Example 2:
the embodiment 2 of the present disclosure provides a readable storage medium, on which a program is stored, and when the program is executed by a processor, the method for filtering solar radio interference according to the embodiment 1 of the present disclosure is implemented.
Example 3:
the embodiment 3 of the present disclosure provides an electronic device, which includes a memory, a processor, and a program stored in the memory and capable of being executed on the processor, where the processor executes the program to implement the method for filtering out solar radio interference according to the embodiment 1 of the present disclosure.
The electronic device described in this embodiment includes, but is not limited to, a computer, a solar radio frequency spectrometer, various microprocessors, a processor, a server, or various solar radio observation devices.
The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure, and various modifications and changes may be made to the present disclosure by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present disclosure should be included in the protection scope of the present disclosure.
Claims (14)
1. A solar radio interference filtering method is characterized by comprising the following steps:
(1) collecting intensity data of a set time period before and after a solar explosion event, and screening out frequency channels where radio station signals meeting conditions are located according to radio current value data of all frequency channels at a certain moment in a non-explosion period;
(2) preprocessing the collected radio station signal data meeting the conditions to establish a data set;
(3) establishing a cyclic neural network, training the cyclic neural network by utilizing the preprocessed radio station signal data, and predicting a signal value of a radio station in a solar burst area by using the trained cyclic neural network;
(4) and according to the linear additivity of the signals, subtracting the signal value of the predicted radio station from the original value of the radio current at the corresponding moment of the burst area to obtain the value of the clean pure solar burst event.
2. The solar radio interference filtering method according to claim 1, wherein in the step (1),
(1-1) drawing radio flow of all frequency points at a certain moment in a non-outbreak period;
(1-2) determining an intensity threshold according to the magnitude of the radio current value of the explosion event;
and (1-3) selecting the frequency channel with the intensity value larger than the intensity threshold value, and storing the intensity value of the selected radio station signal according to the corresponding frequency channel.
3. The method according to claim 1, wherein in step (1), after the station signal to be processed is screened out, a certain frequency channel of the station signal is selected, and the position of the initial moment of the solar radio burst is located according to a curve of the radio flow of the frequency channel changing with time.
4. The solar radio interference filtering method according to claim 1, wherein in the step (2), the acquired data is subjected to FFT operation and digital polarization synthesis operation in the FPGA, and finally a left-hand signal and a right-hand signal are obtained;
reading data according to a predetermined data format and performing accumulation operation within a time resolution, and selecting one of a left-handed signal and a right-handed signal as a signal to be processed.
5. The method of claim 1, wherein the training data is normalized by a min-max normalization method to convert the data linearly to [0, 1 ].
6. The solar radio interference filtering method as claimed in claim 1, wherein the radio signal sequence is segmented in time for establishing a mapping relationship between time segments.
7. The solar radio interference filtering method according to claim 1, wherein in the step (3), pure electric station data before the initial moment of solar burst is firstly divided into a training set and a test set, a network trained by the training set is used for predicting station data and comparing the predicted value with the original pure electric station data at the corresponding moment, and the structure of the cyclic neural network is adjusted and optimized according to the comparison result; and then, all the radio station data before the initial moment of the solar burst are used as a training set, the optimized cyclic neural network structure is trained and adjusted by utilizing the radio station data, and the radio station signal value of the burst area is predicted.
8. The method as claimed in claim 1, wherein in step (3), the fixed several types of values of the radio station of the single frequency channel are digitally mapped by using the distribution characteristic rule of the radio station interference signal values, so that the prediction problem is first converted into a classification problem to be processed to reduce the error caused by the step prediction.
9. The method for filtering solar radio interference according to claim 1, wherein in the step (3), the recurrent neural network includes an input layer, a hidden layer and an output layer, and in the forward transmission process of data, the hidden layer receives not only the information of the input layer at the current moment, but also the information of the hidden layer unit at the previous moment, so as to realize the memory of the information characteristics at the past moment.
10. The method according to claim 1, wherein in the step (3), the recurrent neural network is a long-term and short-term memory network, and an information gate for determining information is added to a module repeatedly linked in a hidden layer of the long-term and short-term memory network, so as to avoid disappearance of gradients caused by too many layers.
11. The method for filtering solar-radio interference as claimed in claim 10, wherein in step (3), the information gate is generated by Sigmoid function, and includes an input gate, a forgetting gate and an output gate, and each information passing through the information gate has a real parameter in the range of [0, 1], and the real parameter represents the proportion of information passing through each gate.
12. The solar radio interference filtering method according to claim 1, wherein in the step (4), the signal values of the non-solar burst area radio station are processed by using a common image filtering method.
13. A readable storage medium, on which a program is stored, which program, when being executed by a processor, is adapted to carry out the steps of the method for solar radio interference rejection as set forth in any one of claims 1 to 12.
14. An electronic device comprising a memory, a processor and a program stored on the memory and executable on the processor, wherein the processor when executing the program performs the steps of the method for filtering solar radio interference according to any of claims 1-12.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910758704.8A CN110632386B (en) | 2019-08-16 | 2019-08-16 | Solar radio interference filtering method, readable storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910758704.8A CN110632386B (en) | 2019-08-16 | 2019-08-16 | Solar radio interference filtering method, readable storage medium and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110632386A CN110632386A (en) | 2019-12-31 |
| CN110632386B true CN110632386B (en) | 2021-07-30 |
Family
ID=68970495
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910758704.8A Expired - Fee Related CN110632386B (en) | 2019-08-16 | 2019-08-16 | Solar radio interference filtering method, readable storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110632386B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113095215B (en) * | 2021-04-09 | 2022-08-26 | 山东大学 | Solar radio filtering method and system based on improved LSTM network |
| CN113804982B (en) * | 2021-09-18 | 2022-06-21 | 山东大学 | Solar radio burst real-time detection method and system based on digital filtering |
| CN115856425B (en) * | 2022-11-21 | 2023-10-17 | 中国人民解放军32802部队 | Spectrum anomaly detection method and device based on hidden space probability prediction |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN205545224U (en) * | 2016-02-03 | 2016-08-31 | 苏州大学 | Signal receiving device of solar radio frequency spectrograph |
| CN106027176A (en) * | 2016-06-29 | 2016-10-12 | 中国人民解放军军械工程学院 | Receiver in-band dual frequency electromagnetic radiation large signal blocking interference effect prediction method |
| CN106124859A (en) * | 2016-08-03 | 2016-11-16 | 山东大学（威海） | Solar radio solar radio burst variable-resolution spectral means based on high-speed AD acquisition card and method thereof |
| CN109063302A (en) * | 2018-07-24 | 2018-12-21 | 山东大学 | The high-resolution solar radio radiation data processing system and method for adaptive outburst judgement |
-
2019
- 2019-08-16 CN CN201910758704.8A patent/CN110632386B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN205545224U (en) * | 2016-02-03 | 2016-08-31 | 苏州大学 | Signal receiving device of solar radio frequency spectrograph |
| CN106027176A (en) * | 2016-06-29 | 2016-10-12 | 中国人民解放军军械工程学院 | Receiver in-band dual frequency electromagnetic radiation large signal blocking interference effect prediction method |
| CN106124859A (en) * | 2016-08-03 | 2016-11-16 | 山东大学（威海） | Solar radio solar radio burst variable-resolution spectral means based on high-speed AD acquisition card and method thereof |
| CN109063302A (en) * | 2018-07-24 | 2018-12-21 | 山东大学 | The high-resolution solar radio radiation data processing system and method for adaptive outburst judgement |
Non-Patent Citations (3)
| Title |
|---|
| Convolutional Neural Network for Classification of Solar Radio Spectrum;Sisi Chen 等;《Proceedings of the IEEE International Conference on Multimedia and Expo Workshops (ICMEW) 2017》;20171231;第198-201页 * |
| Traffic Flow Prediction With Big Data:A Deep Learning Approach;Yisheng Lv 等;《IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS》;20150430;第865-873页 * |
| 太阳射电观测***多通道变频电路一致性补偿方法与实现;杜清府 等;《中国科学：技术科学》;20190814;第49卷(第8期);第901-909页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110632386A (en) | 2019-12-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110632386B (en) | Solar radio interference filtering method, readable storage medium and electronic equipment | |
| Appel et al. | Adaptive sequential segmentation of piecewise stationary time series | |
| CN111193256A (en) | Power load prediction method based on variational modal decomposition and gated cyclic unit | |
| CN108801251B (en) | Inertial sensor aliasing interference signal separation method | |
| CN111340288A (en) | Urban air quality time sequence prediction method considering space-time correlation | |
| CN104168075B (en) | Frequency spectrum detecting method and device in the case of a kind of without knowledge of noise covariance | |
| Knight et al. | Spectral estimation for locally stationary time series with missing observations | |
| CN114626193A (en) | Method for reducing noise of leakage structure vibration signal based on improved variational modal decomposition | |
| CN113723171A (en) | Electroencephalogram signal denoising method based on residual error generation countermeasure network | |
| Mirmomeni et al. | Recursive spectral analysis of natural time series based on eigenvector matrix perturbation for online applications | |
| CN117590142B (en) | Switch cabinet fault diagnosis method and system based on deep learning | |
| CN113095215B (en) | Solar radio filtering method and system based on improved LSTM network | |
| Huang et al. | Application of a novel constrained wavelet threshold denoising method in ensemble-based background-error variance | |
| CN116466316A (en) | Ultra-wideband radar vital sign signal denoising algorithm based on PE and TVF-EMD | |
| CN116155319A (en) | Electromagnetic situation monitoring and analyzing system and method | |
| CN112616160B (en) | Intelligent short-wave frequency cross-frequency-band real-time prediction method and system | |
| Qiang et al. | Detection and mitigation of RFI in SBRS observation data | |
| CN112784686A (en) | Adaptive signal analysis method based on continuous multivariable variational modal decomposition | |
| Matz et al. | Time-frequency methods for signal detection with application to the detection of knock in car engines | |
| CN117454095B (en) | Bridge dynamic deflection data noise reduction method | |
| Sameh et al. | Seasonal Autoregressive Integrated Moving Average for Climate Change Time Series Forecasting | |
| CN116704007B (en) | Sea ice area change statistical method, system, electronic equipment and medium | |
| CN114611329B (en) | Time domain electromagnetic method near field noise suppression method based on variation modal decomposition | |
| CN114255182B (en) | CS iteration threshold image denoising reconstruction method based on space self-adaptive total variation | |
| CN110224767B (en) | Fault prediction method for short wave signaling system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210730 |