CN116996111A - Satellite spectrum prediction method and device and electronic equipment - Google Patents
Satellite spectrum prediction method and device and electronic equipment Download PDFInfo
- Publication number
- CN116996111A CN116996111A CN202311065288.6A CN202311065288A CN116996111A CN 116996111 A CN116996111 A CN 116996111A CN 202311065288 A CN202311065288 A CN 202311065288A CN 116996111 A CN116996111 A CN 116996111A
- Authority
- CN
- China
- Prior art keywords
- spectrum
- data
- satellite
- spectrum data
- band
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001228 spectrum Methods 0.000 title claims abstract description 275
- 238000000034 method Methods 0.000 title claims abstract description 44
- 230000003595 spectral effect Effects 0.000 claims abstract description 46
- 230000007246 mechanism Effects 0.000 claims abstract description 30
- 239000011159 matrix material Substances 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 7
- 238000003064 k means clustering Methods 0.000 claims description 5
- 238000013450 outlier detection Methods 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 238000011524 similarity measure Methods 0.000 claims description 4
- 238000004891 communication Methods 0.000 description 25
- 238000010586 diagram Methods 0.000 description 10
- 230000007774 longterm Effects 0.000 description 9
- 238000012549 training Methods 0.000 description 7
- 238000012986 modification Methods 0.000 description 5
- 230000004048 modification Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000013475 authorization Methods 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000000873 masking effect Effects 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 230000011664 signaling Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000013468 resource allocation Methods 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/14—Relay systems
- H04B7/15—Active relay systems
- H04B7/185—Space-based or airborne stations; Stations for satellite systems
- H04B7/1851—Systems using a satellite or space-based relay
- H04B7/18513—Transmission in a satellite or space-based system
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/382—Monitoring; Testing of propagation channels for resource allocation, admission control or handover
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/391—Modelling the propagation channel
- H04B17/3913—Predictive models, e.g. based on neural network models
Abstract
The application provides a satellite spectrum prediction method, a satellite spectrum prediction device and electronic equipment, wherein the method comprises the following steps: collecting spectrum data of a target satellite, wherein the spectrum data is expressed as a wide-band spectrum power value; clustering the spectrum data, and extracting sub-band spectrum data in the spectrum data; temporally position-coding sub-band spectral data; inputting the coded sub-band spectrum data into a ConvTransformer model, and predicting the spectrum of a target satellite in a period of time, wherein the ConvTransformer model is based on a Transformer architecture and comprises a coding layer and a decoding layer, and the coding layer sequentially comprises a multi-head attention mechanism, a convolution layer and a feedforward network; the encoded sub-band spectral data is first input into a multi-headed attention mechanism. The spectrum prediction method has the advantages of less data operation amount, high operation speed, high prediction accuracy and the like.
Description
Technical Field
The present application relates to the field of satellite communications technologies, and in particular, to a satellite spectrum prediction method, a device, and an electronic device.
Background
Currently, space satellite systems include Geostationary Satellites (GSOs) and non-geostationary satellites (NGSOs) according to orbital altitude partitioning. Because the geostationary satellite is in a special position and is stationary relative to the ground, the geostationary satellite plays a very important role in the fields of geostationary observation, communication, remote sensing, weather, broadcasting and the like. The frequency band of serious co-channel interference of the high-orbit satellite and the low-orbit satellite is mainly indicated in the relevant regulations of the international organizations such as Ku frequency band, ITU and the like, and the non-geostationary satellite cannot interfere with the normal operation of a geostationary satellite system. In the C band, there is also mutual interference between the satellite system and the terrestrial network. The design of the space spectrum sharing system is not only limited by the short-cut space spectrum resources, but also limited by the communication round trip high time delay, and the single-step prediction result is difficult to support the signaling overhead of the spectrum resource allocation in the sharing system.
The spectrum sharing strategy based on cognition has been studied deeply on the ground, and in a space segment satellite spectrum system, satellites with arbitrary orbit heights are observed in real time on the ground to serve as main users (PU), other communication systems which generate same-frequency interference with the main users are analyzed to serve as Secondary Users (SU), and the problem between the main users and the secondary users can be high-orbit and low-orbit constellation spectrum problems or satellite and ground network spectrum sharing problems. And in the satellite shared frequency band, monitoring the spectrum use condition of the primary user in real time, and distributing the future idle frequency band to the secondary user through the centralized analysis of the ground network. Through the satellite spectrum sharing system, the space overall spectrum utilization rate can be improved, and the problem of co-channel interference can be solved.
The accuracy of the prediction mode of the satellite spectrum is low, the satellite-to-ground communication time delay is large, and the practicability is insufficient. For example, in a spectrum prediction method based on a convolutional neural network, training is performed by using the convolutional neural network by sensing a spectrum channel environment, and a free channel with the highest probability in the future is output, and a mechanism based on collision rate is designed. In another example, a certain cross-frequency-band spectrum prediction method based on deep migration learning is used for grasping the similarity of each channel between frequency bands, mining the internal law and the correlation among spectrum data through historical spectrum data measured on channels of other frequency bands, migrating to the current frequency band so as to predict the spectrum state of the current frequency band at the future time, wherein the spectrum prediction model only predicts the channel state at one time in the future, and the single-time-slot prediction in a satellite communication scene cannot solve the satellite-to-ground communication time delay problem.
Therefore, the accuracy of the current satellite spectrum prediction mode is low, the satellite shared spectrum scene cannot be well dealt with, and the satellite-to-ground communication time delay problem cannot be solved.
Disclosure of Invention
The technical problem to be solved by the application is to provide a satellite spectrum prediction method, a satellite spectrum prediction device and electronic equipment, which have the advantages of less data operand, high calculation speed and high prediction accuracy.
In order to solve the above technical problems, in a first aspect, the present application provides a satellite spectrum prediction method, including: collecting spectrum data of a target satellite, wherein the spectrum data is expressed as a wide-band spectrum power value; clustering the spectrum data, and extracting sub-band spectrum data in the spectrum data; temporally position-coding the sub-band spectral data; inputting the coded sub-band spectrum data into a ConvTransformer model, and predicting the spectrum of the target satellite within a period of time; the ConvTransformer model is based on a Transformer architecture and comprises a coding layer and a decoding layer, wherein the coding layer sequentially comprises a multi-head attention mechanism, a convolution layer and a feedforward network; the encoded sub-band spectral data is first input into a multi-head attention mechanism.
Optionally, the acquiring the spectrum data of the target satellite includes: and collecting spectrum power exceeding a certain threshold value in the target satellite as the spectrum data.
Optionally, after collecting the spectrum data of the target satellite, the method further comprises: and storing the acquired spectrum data.
Optionally, storing the collected spectral data includes: the acquired spectral data is stored in the form of a multi-band time series.
Optionally, clustering the spectrum data includes: and clustering the spectrum data by adopting a K-means clustering algorithm.
Optionally, after collecting the spectrum data of the target satellite, before clustering the spectrum data, the method further includes: preprocessing the spectrum data and optimizing the spectrum data.
Optionally, preprocessing the spectral data includes one or more of: outlier detection, missing value filling, and gaussian white noise removal.
Optionally, the convolution layer employs one-dimensional convolution for extracting correlation of adjacent subband spectral data.
Optionally, temporally position encoding the subband spectral data comprises: the position coding is performed alternately by sine and cosine functions:
wherein PE represents position coding, pos represents an index of a position where current data is located, i represents an index of a dimension where current time is located, d mdl Representing the overall number of dimensions of the position code.
Optionally, in the multi-head attention mechanism and the second multi-head attention mechanism, the attention mechanism calculating method is as follows:
wherein Q is a query matrix, represents currently input data to be matched, K is a key value matrix, represents other data to be matched, V is a value matrix, represents the size of the data, and d k Representing normalization of the data, softmax () is an activation function, mapping the input value between 0 and 1, attention (Q, K, V) represents the Attention score, which is a criterion for the similarity measure.
In a second aspect, the present application provides a satellite spectrum prediction apparatus, comprising: the acquisition module is used for acquiring spectrum data of a target satellite, wherein the spectrum data are expressed as wide-band spectrum power values; the clustering module is used for carrying out clustering processing on the spectrum data and extracting sub-band spectrum data in the spectrum data; the coding module is used for performing position coding on the sub-band spectrum data in time; the prediction module is used for inputting the coded sub-band spectrum data into a ConvTransformer model and predicting the spectrum of the target satellite in a period of time; the ConvTransformer model is based on a Transformer architecture and comprises a coding layer and a decoding layer, wherein the coding layer sequentially comprises a multi-head attention mechanism, a convolution layer and a feedforward network; the encoded sub-band spectral data is first input into a multi-head attention mechanism.
In a third aspect, the present application provides an electronic device, comprising: a processor and a memory storing a program or instructions executable on the processor, which when executed by the processor, implement the steps of the satellite spectrum prediction method as described in the first aspect.
In a fourth aspect, the present application provides a readable storage medium having stored thereon a program or instructions which when executed by a processor performs the steps of the satellite spectrum prediction method according to the first aspect.
Compared with the prior art, the application has the following advantages: firstly, collecting spectrum data of a target satellite, wherein the spectrum data is expressed as a wide-band spectrum power value; clustering the spectrum data to extract sub-band spectrum data in the spectrum data; position coding is carried out on sub-band spectrum data in time; and finally, inputting the coded sub-band spectrum data into a ConvTransformer model, predicting the spectrum of a target satellite in a period of time, and on the premise of ensuring the prediction accuracy, the complexity of receiving the satellite spectrum data can be reduced well and the satellite spectrum prediction result can be calculated rapidly.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the principles of the application. In the accompanying drawings:
FIG. 1 is a flowchart of a satellite spectrum prediction method according to an embodiment of the present application;
FIG. 2 is a schematic diagram of a manner in which satellite spectrum prediction may be implemented in accordance with an embodiment of the present application;
FIG. 3 is a diagram showing short-term prediction results according to an embodiment of the present application;
FIG. 4 is a diagram showing long-term prediction results according to an embodiment of the present application;
FIG. 5 is a schematic diagram of a satellite spectrum prediction apparatus according to an embodiment of the present application;
fig. 6 is a schematic diagram of an electronic device according to an embodiment of the application.
Detailed Description
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are used in the description of the embodiments will be briefly described below. It is apparent that the drawings in the following description are only some examples or embodiments of the present application, and it is apparent to those of ordinary skill in the art that the present application may be applied to other similar situations according to the drawings without inventive effort. Unless otherwise apparent from the context of the language or otherwise specified, like reference numerals in the figures refer to like structures or operations.
As used in the specification and in the claims, the terms "a," "an," "the," and/or "the" are not specific to a singular, but may include a plurality, unless the context clearly dictates otherwise. In general, the terms "comprises" and "comprising" merely indicate that the steps and elements are explicitly identified, and they do not constitute an exclusive list, as other steps or elements may be included in a method or apparatus.
A flowchart is used in the present application to describe the operations performed by a system according to embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in order precisely. Rather, the various steps may be processed in reverse order or simultaneously. At the same time, other operations are added to or removed from these processes.
Example 1
Fig. 1 is a flowchart of a satellite spectrum prediction method according to an embodiment of the application, and referring to fig. 1, a method 100 includes: s110, collecting spectrum data of a target satellite, wherein the spectrum data are expressed as wide-band spectrum power values; s120, clustering the spectrum data, and extracting sub-band spectrum data in the spectrum data; s130, performing position coding on the sub-band spectrum data in time; s140, inputting the encoded sub-band spectrum data into a ConvTransformer model, and predicting the spectrum of the target satellite in a period of time. The ConvTransformer model is based on a Transformer architecture and comprises an encoding layer and a decoding layer, wherein the encoding layer sequentially comprises a multi-head attention mechanism, a convolution layer and a feedforward network, and the encoded sub-band spectrum data is firstly input into the multi-head attention mechanism.
Current satellite communication systems often require spectrum sharing. For example, based on a satellite system including a primary user satellite (pu_sat) and a secondary user satellite (su_sat), the primary user satellite and the secondary user satellite have overlapping spectrums within authorized available spectrum resources, if co-channel interference problems occur when using the same frequency band resources, and when co-channel interference occurs between the primary user satellite and the secondary user satellite, the secondary user satellite needs to avoid the primary user satellite. The primary user satellite generally refers to a satellite with higher priority, preferably occupies spectrum resources, and may be a high-orbit satellite (GEO), a medium-orbit satellite (MEO) or a low-orbit satellite (LEO), at least one of which is included, and adopts a frequency division multiplexing mode. The secondary user satellite can be a satellite system or a ground wireless network, at least one secondary user satellite is/are contained, and a frequency division multiplexing mode is adopted. Therefore, to realize high-low orbit satellite spectrum sharing, it is necessary to improve the accuracy of multi-step satellite spectrum, and reduce spectrum conflicts caused by communication delay of heterogeneous satellite systems according to the prediction result, especially the long-term prediction result.
In order to solve the problem of satellite spectrum interference, the method of the embodiment firstly predicts the spectrum of a target satellite (such as a main user satellite) to obtain spectrum data in a future period, so that spectrum authorization can be carried out on a secondary user satellite according to the predicted spectrum data to complete tasks such as satellite spectrum sharing, interference avoidance and the like. Meanwhile, the method of the embodiment can predict the frequency spectrum of a period of time in the future, so that the frequency spectrum conflict caused by the communication time delay of the heterogeneous satellite system can be reduced. In addition, the method adopts clustering processing, introduces satellite spectrum clustering processing in advance, clusters spectrum data into sub-band spectrum data, and can better reduce complexity of collecting satellite spectrum data.
Broadband multi-step satellite spectrum prediction places high demands on computational complexity and prediction accuracy. In this embodiment, a conv transducer model (improved transducer model) is adopted, which is based on a traditional transducer timing prediction architecture on one hand, and a convolutional layer is added to a coding layer on the other hand, so as to extract relevant information of adjacent sub-bands, thereby improving spectrum prediction accuracy. The transducer model is a neural network model based on an attention mechanism, which has better parallelization capability and shorter training time and is excellent in processing long-sequence tasks, compared to the conventional cyclic neural network model (e.g., LSTM and GRU).
Specifically, the method of the embodiment can perform clustering processing on satellite frequency point data acquired by a spectrum receiver, output sub-band spectrum data, retain the time sequence of the satellite frequency spectrum data in the clustering process, take the average power value of frequency points in each sub-band, and perform position coding, wherein the design of the position coding accords with the time slot of satellite transmission data, and is more accurate. The timing prediction is based in part on the encoder and decoder architecture of the transducer architecture. In the coding layer, in order to analyze the characteristics of the subband spectrum data, a convolution layer (Conv layer) is designed, a convolution window slides along a frequency domain axis, the subband spectrum characteristics are extracted, and compared with a traditional transform model, the overall prediction accuracy of a long-term satellite spectrum is further improved.
In an example, the collecting the spectral data of the target satellite may be collecting the spectral power exceeding a certain threshold value in the target satellite as the spectral data. For example, when the main user satellite performs communication data transmission, the received spectrum power exceeds a certain threshold value to indicate that the frequency band is in an occupied state at the current moment, otherwise, the frequency band is in an idle state, and the embodiment takes the spectrum power in the occupied state as spectrum data, so that the spectrum data condition of the target satellite can be reflected better.
In one example, after the spectral data of the target satellite is acquired, the acquired spectral data is stored. Further, the acquired spectral data is stored in a multi-band time series.
Illustratively, the spectrum data is collected by a Radio Monitoring Center (RMC), which may be configured with a high gain antenna, directed to the primary user satellite, and receives signals from the primary user satellite in real time, which may be satellite signals received in real time by a multi-beam antenna, and the spectrum sharing system is downlink. When the spectrum data is required to be stored, the radio monitoring center monitors the main user satellite signals above the deployment area in real time, receives the data, records the spectrum resource use condition of the main user satellite, and stores the spectrum resource use condition into a spectrum resource database. In one implementation, the radio monitoring center receives data from the user satellite as wide-band spectral power values sampled by the spectrometer and stored locally in time sequence.
In one example, clustering the spectral data may be clustering the spectral data using a K-means clustering algorithm. The wideband spectral data is clustered to analyze which sub-band spectral data is occupied by the primary user satellite communications.
In this embodiment, the ConvTransformer model has a high requirement on the calculation amount, if all the spectrum data received by the gateway station are input into the spectrum prediction model, the calculation amount is huge, and the time consumption is longer, so before entering the spectrum prediction module, the satellite spectrum data can be clustered in advance, the spectrum data is clustered into sub-band spectrum data, and the complexity of receiving the satellite spectrum data can be reduced well on the premise of ensuring the prediction accuracy. The satellite spectrum clustering uses a clustering algorithm based on K-means, bandwidth frequency point data received by a radio monitoring center can be clustered, and sub-channel frequency power values (in units of dbm or db) in a satellite authorized bandwidth are output. The clustering algorithm can learn the number of sub-band spectrums in the spectrum data received by the radio monitoring center in a training stage to obtain an optimal clustering cluster.
Exemplary, based on K-means clustering algorithm, first obtain wideband frequency band length, define cluster loss functionWherein Loss is c The size of the clustering loss value is that N is the length of a broadband frequency band, G is the number of clusters and x is j Is the size, mu, of the data on the jth frequency point g Is the size of the g-th cluster center; secondly, solving the sum of all frequency point data and clustering center data in Euclidean distance, and iteratively solving the contour coefficient of the center frequency point closest to the center frequency pointWhere a (τ) represents the average distance between a data point and other data points in the same cluster, b (τ) represents the average distance between a data point and all data points in other clusters, S (τ) represents the contour coefficient, and the point with the smallest overall center clusterS represents the overall minimum contour coefficient after clustering, and N is the length of the broadband frequency band; finally, outputting the optimal sub-channel clustering result +.>Wherein t represents the time t, sigma represents the number of sub-band spectrums output after spectrum clustering, and the output sub-channel spectrum data and the original received data are on a time axisAnd keep the same.
In an example, after collecting the spectrum data of the target satellite, the spectrum data may be preprocessed to optimize the spectrum data before clustering the spectrum data. Further, preprocessing the spectral data includes one or more of the following: outlier detection, missing value filling, and gaussian white noise removal.
Exemplary, the spectrum data of the primary user satellite is collected in real time, and the data is stored as a multi-band time sequenceWherein t represents the t moment, ω represents the ω -th frequency point, x t,ω The power level at the ω -th frequency point at time t is shown. Preprocessing the data, such as outlier detection, missing value filling, and Gaussian white noise removal, to obtain data +.>Wherein->And the power on the omega frequency point at the t time after pretreatment is shown. And further, the spectrum data is optimized, so that the spectrum data can reflect the actual situation of the spectrum of the target satellite.
In one example, the convolution layer employs one-dimensional convolution (conv 1 d) for extracting correlation for adjacent subband spectral data, which may improve prediction accuracy.
The one-dimensional convolution layer is added on the encoder layer, the convolution kernel window slides along the frequency domain direction, the frequency domain characteristics are extracted, the correlation of the sub-band spectrum data is further obtained, and the convolution calculation formula is as follows:
wherein Γ represents half of the subband spectrum length, C represents the input spectrum sequence, ψ is the convolution kernel, ζ is the convolution output, η is the spectrum sequence index, and f is the sliding step size.
In one example, the position encoding of the subband spectral data in time may be by alternating a sine function with a cosine function. The position code is matched with a satellite communication protocol, and a more proper position code processing result is obtained, as follows:
wherein PE represents position coding, pos represents an index of a position where current data is located, i represents an index of a dimension where current time is located, d mdl Representing the overall number of dimensions of the position code. The position coding mode not only keeps the time sequence periodicity of satellite spectrum data, but also keeps the adjacent data difference, and can respectively and alternately code year, month, day, time, minute and second according to the satellite spectrum data.
In one example, in a multi-headed attention mechanism, the attention mechanism calculation method is:wherein Q is a query matrix, represents currently input data to be matched, K is a key value matrix, represents other data to be matched, V is a value matrix, represents the size of the data, and d k Representing normalization of the data, softmax () is an activation function, mapping the input value between 0 and 1, attention (Q, K, V) represents the Attention score, which is a criterion for the similarity measure.
Further, the long-range dependence in the history spectrum data is analyzed by multiplying the input long-sequence satellite spectrum data X by Q, K, V, respectively. Defining a predictive loss functionWherein Loss is p Indicating the magnitude of the predicted loss value, Y m,r Indicating the m-th time, the predicted value size of the r-th subband, < >>Representing the true of the mth time and the nth subbandData value size. The training set data is loaded into a spectrum prediction network for iterative fitting, and data t with fixed length L is randomly intercepted from the training set in each training 1 ,t 2 ,…,t L ]Inputting a spectrum prediction model, fitting, and outputting [ t ] with the prediction length of P L+1 ,t L+2 ,…,t L+P ]And predicting a result.
Fig. 2 is a schematic diagram of a manner of implementing satellite spectrum prediction in an embodiment of the present application, referring to fig. 2, first, input spectrum data is clustered, a large number of frequency points are clustered into spectrum subband data, the subband spectrum data after the clustering is input into a satellite spectrum prediction model, satellite spectrum, especially long-term satellite spectrum prediction is performed, and a spectrum prediction result in a future period is output.
Referring to fig. 2, a Position encoder (Position encoder) designs a time stamp matching a satellite received signal, and encodes the time stamp from year, month, day, and day to second with sine and cosine, respectively, and the encoding has periodicity while retaining a relative Position difference.
Referring to fig. 2, the encoding layer includes a multi-head attention mechanism module, a convolution layer and a feedforward network module, the spectrum data after position encoding is firstly input into the multi-head attention mechanism module, the correlation between the data to be predicted and the history data at different moments is extracted, the correlation is higher, and the attention score is larger.
Referring to fig. 2, the decoding layer includes a masking multi-head attention mechanism, a multi-head attention mechanism and a feed-forward network, wherein the masking multi-head attention mechanism module masks information to be predicted in the future, assigns a negative infinity, and can ensure that real spectrum data in the future can not be acquired in the learning process and correct training is ensured.
Fig. 3 is a view showing a short-term prediction result in an embodiment of the present application, and referring to fig. 3, the size of spectrum data of 10 times in the future is predicted by using spectrum data of 10 times in the history, wherein an abscissa represents a time axis, an ordinate represents an amplitude, a broken line represents a predicted value, prediction is performed from the 11 th time according to the obtained spectrum data of 10 times in the history, and it can be known from the figure that an accurate prediction result has a better result in terms of trend and peak reaching rate, and can have higher accuracy in short-term spectrum prediction.
Fig. 4 is a diagram showing a long-term prediction result in an embodiment of the present application, and referring to fig. 4, spectrum data of 96 times is used to predict spectrum sizes of 96 times in the future, a single spectrum subband in the prediction result is taken for display, and the horizontal axis coordinates respectively represent time and numerical values, and the prediction is performed from 97 th data, and it can be known from the diagram that the long-term spectrum change is fast, the jitter is high, the prediction difficulty is large, the prediction value in the spectrum result is substantially consistent with the true value, the trend is high, the power peak prediction accuracy is high, and the model can be found to have high prediction accuracy, and has great value in practical application.
In this embodiment, the processed subband spectrum data is predicted by a prediction model, the spectrum power in a period of time in the future is output, and the model output result can be a long-term sequence.
After the spectrum prediction is completed, the spectrum prediction data can be applied, for example, a Network Management Center (NMC) performs spectrum authorization on the secondary user satellite according to the spectrum prediction result, so as to complete satellite spectrum sharing, and a section of protection time slot is set according to the satellite downlink communication time delay of the primary user satellite and the satellite signaling authorization uplink and downlink communication time delay of the secondary user, so that the heterogeneous system is distributed in the time dimension. And allocating the idle frequency band resources in the predicted result after the protection time slot. The secondary user satellite can use the frequency band for communication after obtaining the frequency authorization of the network management center. The radio monitoring center can collect the satellite spectrum state of the main user in real time, and when the satellite spectrum data of the main user received by the radio monitoring center is different from the predicted value, the network management center immediately sends an instruction to interrupt the use of the frequency band by the secondary user in the overlapped frequency band.
The satellite spectrum prediction method provided by the embodiment firstly collects the spectrum data of the target satellite, wherein the spectrum data is expressed as a wide-band spectrum power value; clustering the spectrum data to extract sub-band spectrum data in the spectrum data; position coding is carried out on sub-band spectrum data in time; and finally, inputting the coded sub-band spectrum data into a ConvTransformer model, predicting the spectrum of a target satellite in a period of time, and on the premise of guaranteeing the accuracy of spectrum prediction, particularly long-term spectrum prediction, the complexity of receiving the satellite spectrum data can be reduced well and the satellite spectrum prediction result can be calculated rapidly.
Example two
Fig. 5 is a schematic structural diagram of a satellite spectrum prediction apparatus according to an embodiment of the present application, and referring to fig. 5, an apparatus 500 is shown mainly including:
an acquisition module 501, configured to acquire spectrum data of a target satellite, where the spectrum data represents a wide-band spectrum power value; the clustering module 502 is configured to perform clustering processing on the spectrum data, and extract subband spectrum data in the spectrum data; an encoding module 503, configured to temporally perform position encoding on the subband spectrum data; and a prediction module 504, configured to input the encoded subband spectrum data into a conv transducer model, and predict a spectrum of the target satellite in a period of time. The ConvTransformer model is based on a Transformer architecture and comprises a coding layer and a decoding layer, wherein the coding layer sequentially comprises a multi-head attention mechanism, a convolution layer and a feedforward network; the encoded sub-band spectral data is first input into a multi-headed attention mechanism.
In one example, collecting spectral data of the target satellite may include collecting spectral power in the target satellite that exceeds a certain threshold as spectral data.
In one example, after the spectral data of the target satellite is acquired, the acquired spectral data is stored.
In one example, storing the acquired spectral data may include storing the acquired spectral data in a multi-band time series.
In one example, clustering the spectral data includes clustering the spectral data using a K-means clustering algorithm.
In one example, after collecting the spectral data of the target satellite, the spectral data is preprocessed to optimize the spectral data before clustering the spectral data.
In an example, preprocessing the spectral data includes one or more of the following: outlier detection, missing value filling, and gaussian white noise removal.
In one example, the convolution layer employs one-dimensional convolution for extracting correlations for adjacent subband spectral data.
In one example, the position encoding of the subband spectral data in time may be by alternating a sine function with a cosine function:
wherein PE represents position coding, pos represents an index of a position where current data is located, i represents an index of a dimension where current time is located, d mdl Representing the overall number of dimensions of the position code.
In one example, in a multi-headed attention mechanism, the attention mechanism calculation method is:wherein Q is a query matrix, represents currently input data to be matched, K is a key value matrix, represents other data to be matched, V is a value matrix, represents the size of the data, and d k Representing normalization of the data, softmax () is a commonly used activation function, mapping the input values between 0 and 1, attention (Q, K, V) representing the Attention score, which is a criterion for similarity measure.
Reference may be made to the foregoing embodiments for details of other operations performed by the modules in this embodiment, which are not further described herein.
The satellite spectrum prediction device provided by the embodiment firstly collects spectrum data of a target satellite, wherein the spectrum data is expressed as a wide-band spectrum power value; clustering the spectrum data to extract sub-band spectrum data in the spectrum data; position coding is carried out on sub-band spectrum data in time; and finally, inputting the coded sub-band spectrum data into a ConvTransformer model, predicting the spectrum of a target satellite in a period of time, and on the premise of guaranteeing the accuracy of spectrum prediction, particularly long-term spectrum prediction, the complexity of receiving the satellite spectrum data can be reduced well and the satellite spectrum prediction result can be calculated rapidly.
The satellite spectrum prediction device in the embodiment of the application can be a device, and can also be a component, an integrated circuit or a chip in a terminal. A satellite spectrum prediction apparatus in an embodiment of the present application may be an apparatus having an operating system. The operating system may be an android operating system, an iOS operating system, or other possible operating systems, and the embodiment of the present application is not limited specifically.
The application also provides an electronic device, comprising: a memory for storing programs or instructions executable by the processor; and a processor, configured to execute the program or instructions to implement each process of the satellite spectrum prediction method embodiment, and achieve the same technical effects, so that repetition is avoided, and no further description is given here.
Fig. 6 is a schematic diagram of an electronic device according to an embodiment of the application. The electronic device 600 may include an internal communication bus 601, a Processor (Processor) 602, a Read Only Memory (ROM) 603, a Random Access Memory (RAM) 604, and a communication port 605. When applied to a personal computer, the electronic device 600 may also include a hard disk 606. The internal communication bus 601 may enable data communication among the components of the electronic device 600. The processor 602 may make the determination and issue the prompt. In some implementations, the processor 602 may be comprised of one or more processors. The communication port 605 may enable the electronic device 600 to communicate data with the outside. In some implementations, the electronic device 600 may send and receive information and data from a network through the communication port 605. The electronic device 600 may also include various forms of program storage elements and data storage elements such as hard disk 606, read Only Memory (ROM) 603 and Random Access Memory (RAM) 604, capable of storing various data files for computer processing and/or communication, as well as possible programs or instructions for execution by the processor 602. The results processed by the processor 602 are communicated to the user device via the communication port 605 for display on a user interface.
The embodiment of the application also provides a readable storage medium, and the readable storage medium stores a program or an instruction, which when executed by a processor, implements each process of the satellite spectrum prediction method embodiment, and can achieve the same technical effect, so that repetition is avoided, and no further description is provided here.
It will be apparent to those skilled in the art from this disclosure that the foregoing disclosure is by way of example only and is not intended to be limiting. Although not explicitly described herein, various modifications, improvements and adaptations of the application may occur to one skilled in the art. Such modifications, improvements, and modifications are intended to be suggested within the present disclosure, and therefore, such modifications, improvements, and adaptations are intended to be within the spirit and scope of the exemplary embodiments of the present disclosure.
Meanwhile, the present application uses specific words to describe embodiments of the present application. Reference to "one embodiment," "an embodiment," and/or "some embodiments" means that a particular feature, structure, or characteristic is associated with at least one embodiment of the application. Thus, it should be emphasized and should be appreciated that two or more references to "an embodiment" or "one embodiment" or "an alternative embodiment" in various positions in this specification are not necessarily referring to the same embodiment. Furthermore, certain features, structures, or characteristics of one or more embodiments of the application may be combined as suitable.
Similarly, it should be noted that in order to simplify the description of the present disclosure and thereby aid in understanding one or more inventive embodiments, various features are sometimes grouped together in a single embodiment, figure, or description thereof. This method of disclosure, however, is not intended to imply that more features than are required by the subject application. Indeed, less than all of the features of a single embodiment disclosed above.
While the application has been described with reference to the specific embodiments presently, it will be appreciated by those skilled in the art that the foregoing embodiments are merely illustrative of the application, and various equivalent changes and substitutions may be made without departing from the spirit of the application, and therefore, all changes and modifications to the embodiments are intended to be within the scope of the appended claims.
Claims (13)
1. A method for predicting satellite spectrum, comprising:
collecting spectrum data of a target satellite, wherein the spectrum data is expressed as a wide-band spectrum power value;
clustering the spectrum data, and extracting sub-band spectrum data in the spectrum data;
temporally position-coding the sub-band spectral data;
inputting the coded sub-band spectrum data into a ConvTransformer model, and predicting the spectrum of the target satellite within a period of time;
the ConvTransformer model is based on a Transformer architecture and comprises a coding layer and a decoding layer, wherein the coding layer sequentially comprises a multi-head attention mechanism, a convolution layer and a feedforward network; the encoded sub-band spectral data is first input into a multi-head attention mechanism.
2. The satellite spectrum prediction method of claim 1 wherein said acquiring the spectrum data of the target satellite comprises: and collecting spectrum power exceeding a certain threshold value in the target satellite as the spectrum data.
3. The satellite spectrum prediction method of claim 1, further comprising, after collecting the spectrum data of the target satellite: and storing the acquired spectrum data.
4. The satellite spectrum prediction method of claim 3 wherein storing the collected spectrum data comprises: the acquired spectral data is stored in the form of a multi-band time series.
5. The satellite spectrum prediction method of claim 1 wherein clustering the spectral data comprises: and clustering the spectrum data by adopting a K-means clustering algorithm.
6. The satellite spectrum prediction method of claim 1 wherein after collecting the spectrum data of the target satellite, before clustering the spectrum data, further comprising: preprocessing the spectrum data and optimizing the spectrum data.
7. The satellite spectrum prediction method of claim 6 wherein preprocessing the spectrum data comprises one or more of: outlier detection, missing value filling, and gaussian white noise removal.
8. The satellite spectrum prediction method of claim 1 wherein said convolution layer employs one-dimensional convolution for extracting correlations for adjacent said sub-band spectrum data.
9. The satellite spectrum prediction method of claim 1 wherein temporally position encoding the sub-band spectrum data comprises:
the position coding is performed alternately by sine and cosine functions:
wherein PE represents position coding, pos represents an index of a position where current data is located, i represents an index of a dimension where current time is located, d mdl Representing the overall number of dimensions of the position code.
10. The satellite spectrum prediction method according to claim 1, wherein in the multi-head attention mechanism, an attention mechanism calculation method is as follows:
wherein Q is a query matrix, represents currently input data to be matched, K is a key value matrix, represents other data to be matched, V is a value matrix, represents the size of the data, and d k Representing normalization of the data, softmax () is an activation function, mapping the input value between 0 and 1, attention (Q, K, V) represents the Attention score, which is a criterion for the similarity measure.
11. A satellite spectrum prediction apparatus, comprising:
the acquisition module is used for acquiring spectrum data of a target satellite, wherein the spectrum data are expressed as wide-band spectrum power values;
the clustering module is used for carrying out clustering processing on the spectrum data and extracting sub-band spectrum data in the spectrum data;
the coding module is used for performing position coding on the sub-band spectrum data in time;
the prediction module is used for inputting the coded sub-band spectrum data into a ConvTransformer model and predicting the spectrum of the target satellite in a period of time;
wherein the ConvTransformer model is based on a Transformer architecture and comprises an encoding layer and a decoding layer, wherein the encoding layer sequentially comprises a multi-head attention mechanism, a convolution layer and a feedforward network; the encoded sub-band spectral data is first input into a multi-head attention mechanism.
12. An electronic device, comprising: a processor and a memory storing a program or instructions executable on the processor, which when executed by the processor, implement the steps of the satellite spectrum prediction method as claimed in any one of claims 1 to 10.
13. A readable storage medium, characterized in that the readable storage medium has stored thereon a program or instructions which, when executed by a processor, implement the steps of the satellite spectrum prediction method according to any of claims 1-10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311065288.6A CN116996111A (en) | 2023-08-23 | 2023-08-23 | Satellite spectrum prediction method and device and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311065288.6A CN116996111A (en) | 2023-08-23 | 2023-08-23 | Satellite spectrum prediction method and device and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116996111A true CN116996111A (en) | 2023-11-03 |
Family
ID=88521318
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311065288.6A Pending CN116996111A (en) | 2023-08-23 | 2023-08-23 | Satellite spectrum prediction method and device and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116996111A (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107359948A (en) * | 2017-07-11 | 2017-11-17 | 北京邮电大学 | A kind of spectrum prediction method and device of cognition wireless network |
| CN114218870A (en) * | 2021-12-22 | 2022-03-22 | 大连理工大学 | Wind speed prediction method based on variational modal decomposition and attention mechanism |
| CN115345220A (en) * | 2022-07-21 | 2022-11-15 | 深兰人工智能科学技术研究院(上海)有限公司 | Prediction method, device and system based on multivariate time sequence |
| CN115694692A (en) * | 2021-12-28 | 2023-02-03 | 南京工业大学 | Spectrum sensing method based on Transformer multi-head attention mechanism network under Alpha noise |
| CN115687989A (en) * | 2022-07-04 | 2023-02-03 | 南京工业大学 | Multi-user spectrum sensing method based on Transformer-DNN hybrid network under Laplace noise |
| WO2023052827A1 (en) * | 2021-10-01 | 2023-04-06 | Telefonaktiebolaget Lm Ericsson (Publ) | Processing a sequence of data items |
| US20230194689A1 (en) * | 2021-12-22 | 2023-06-22 | Battelle Memorial Institute | System and method to extract satellite doppler curves from waterfall spectrograms data |
| CN116405139A (en) * | 2023-03-09 | 2023-07-07 | 西安电子科技大学 | Spectrum prediction model and method based on Informar |
-
2023
- 2023-08-23 CN CN202311065288.6A patent/CN116996111A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107359948A (en) * | 2017-07-11 | 2017-11-17 | 北京邮电大学 | A kind of spectrum prediction method and device of cognition wireless network |
| WO2023052827A1 (en) * | 2021-10-01 | 2023-04-06 | Telefonaktiebolaget Lm Ericsson (Publ) | Processing a sequence of data items |
| CN114218870A (en) * | 2021-12-22 | 2022-03-22 | 大连理工大学 | Wind speed prediction method based on variational modal decomposition and attention mechanism |
| US20230194689A1 (en) * | 2021-12-22 | 2023-06-22 | Battelle Memorial Institute | System and method to extract satellite doppler curves from waterfall spectrograms data |
| CN115694692A (en) * | 2021-12-28 | 2023-02-03 | 南京工业大学 | Spectrum sensing method based on Transformer multi-head attention mechanism network under Alpha noise |
| CN115687989A (en) * | 2022-07-04 | 2023-02-03 | 南京工业大学 | Multi-user spectrum sensing method based on Transformer-DNN hybrid network under Laplace noise |
| CN115345220A (en) * | 2022-07-21 | 2022-11-15 | 深兰人工智能科学技术研究院(上海)有限公司 | Prediction method, device and system based on multivariate time sequence |
| CN116405139A (en) * | 2023-03-09 | 2023-07-07 | 西安电子科技大学 | Spectrum prediction model and method based on Informar |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10674434B2 (en) | Frequency spectrum prediction method and apparatus for cognitive wireless network | |
| Cohen et al. | Spectrum sharing radar: Coexistence via Xampling | |
| US8761677B2 (en) | Multiple stage hybrid spectrum sensing methods and systems for cognitive radio | |
| CN110320535B (en) | Satellite navigation receiver multi-interference suppression method based on wavelet packet transformation and airspace processing | |
| CN101630982A (en) | Broadband cognitive radio frequency spectrum detection method based on parallelly compressed sensing | |
| CN113242557B (en) | Deep learning method and system for spectrum sharing of partially overlapped channels | |
| Kakalou et al. | A survey on spectrum sensing algorithms for cognitive radio networks | |
| Sheikh et al. | Cognitive spectrum sensing and detection using polyphase DFT filter banks | |
| Stinco et al. | Compressed spectrum sensing in cognitive radar systems | |
| Gorcin et al. | Signal identification for adaptive spectrum hyperspace access in wireless communications systems | |
| CN113485212A (en) | Broadband satellite signal intelligent identification system | |
| US6430239B1 (en) | Process of cyclic detection in diversity of polarization of digital cyclostationary radioelectric signals | |
| EP2086255B1 (en) | Process for sensing vacant sub-space over the spectrum bandwidth and apparatus for performing the same | |
| CN116996111A (en) | Satellite spectrum prediction method and device and electronic equipment | |
| CN109756283B (en) | Spectrum sensing method, device and medium for downlink of GEO satellite communication system | |
| CN116321183A (en) | Method, system and electronic equipment for coordinating interference | |
| Liu et al. | Incremental learning based radio frequency fingerprint identification using intelligent representation | |
| Rau et al. | Rfi mitigation for the ngvla: A cost-benefit analysis ngvla memo# 70 | |
| Shen et al. | Deep learning based source number estimation with single-channel mixtures | |
| EP1584943B1 (en) | High resolution direction finding method with broad band FFT sensors | |
| Balint et al. | Application of a probabilistic model to high frequency spectrum occupancy | |
| Balint et al. | Markov model for HF spectrum occupancy | |
| Mariani et al. | Recent advances on wideband spectrum sensing for cognitive radio | |
| Daoud et al. | CNN-Based on-Board Interference Detection in Satellite Systems: AN Analysis of Dataset Impact on Performance | |
| US11108457B2 (en) | Spatial energy rank detector and high-speed alarm |
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 |