CN115577253B - Supervision spectrum sensing method based on geometric power - Google Patents

Supervision spectrum sensing method based on geometric power Download PDF

Info

Publication number
CN115577253B
CN115577253B CN202211470072.3A CN202211470072A CN115577253B CN 115577253 B CN115577253 B CN 115577253B CN 202211470072 A CN202211470072 A CN 202211470072A CN 115577253 B CN115577253 B CN 115577253B
Authority
CN
China
Prior art keywords
learning model
supervised learning
geometric power
noise
model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202211470072.3A
Other languages
Chinese (zh)
Other versions
CN115577253A (en
Inventor
骆忠强
胡倩
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Sichuan University of Science and Engineering
Original Assignee
Sichuan University of Science and Engineering
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Sichuan University of Science and Engineering filed Critical Sichuan University of Science and Engineering
Priority to CN202211470072.3A priority Critical patent/CN115577253B/en
Publication of CN115577253A publication Critical patent/CN115577253A/en
Application granted granted Critical
Publication of CN115577253B publication Critical patent/CN115577253B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Monitoring And Testing Of Transmission In General (AREA)

Abstract

The invention discloses a supervision spectrum sensing method based on geometric power, which relates to the field of spectrum sensing and comprises the following steps: constructing a noise model with generalized Gaussian distribution, and simulating and receiving signals in a noise environment generated by the noise model; carrying out geometric power solution on the received signals, and taking the obtained geometric power as a feature vector; constructing a supervised learning model, and training the supervised learning model through the feature vectors to obtain a trained supervised learning model; and acquiring and inputting the geometric power of the signal in the actual environment into the trained supervised learning model for spectrum sensing. The method has better sensing accuracy rate compared with a mode of sensing the frequency spectrum by using energy statistics and differential entropy by acquiring the geometric power of the received signal as the basis for sensing whether the frequency spectrum is used, especially under the condition of low signal-to-noise ratio.

Description

Supervision spectrum sensing method based on geometric power
Technical Field
The invention relates to the field of spectrum sensing, in particular to a supervision spectrum sensing method based on geometric power.
Background
With the rapid development of wireless communication technology, more and more intelligent devices begin to become a part of internet of things and wireless communication, the demand of the devices on frequency spectrum is very large, and frequency spectrum resources are limited, so how to improve the frequency spectrum utilization rate is the focus of research now, and Cognitive Radio (CR) is one of the methods for solving the problem. Since Primary Users (PUs), i.e., users having licensed spectrum, are not using spectrum at all times, when a PU does not use the licensed spectrum, CR uses opportunity to use idle spectrum, so CR is also called Secondary Users (SUs), and SU can only use spectrum when PU is inactive, once PU uses spectrum, SU must quit using immediately and ensure no interference to PU, so it is very critical to detect whether spectrum is used by PU timely and accurately, spectrum Sensing (SS) in CR can solve this problem, but since wireless communication environment is very complex, SS process is affected by noise, shadow, multipath effect and other problems, resulting in poor detection performance, there are many studies on how to improve SS detection performance, especially under low signal-to-noise ratio.
Disclosure of Invention
Aiming at the defects in the prior art, the supervision spectrum sensing method based on the geometric power solves the problem that the existing spectrum sensing method is low in detection accuracy under the condition of low signal to noise ratio.
In order to achieve the purpose of the invention, the invention adopts the technical scheme that:
a supervised spectrum sensing method based on geometric power is provided, which comprises the following steps:
s1, constructing a noise model with generalized Gaussian distribution, and simulating and receiving signals in a noise environment generated by the noise model;
s2, solving the geometric power of the received signal, and taking the obtained geometric power as a feature vector;
s3, constructing a supervised learning model, and training the supervised learning model through the feature vectors to obtain a trained supervised learning model;
and S4, acquiring and inputting the geometric power of the signal in the actual environment into the trained supervised learning model for spectrum sensing.
Further, in step S1, the shape parameter of the noise is greater than 0 and less than or equal to 2, and the scale parameter of the noise is greater than 0.
Further, the specific method of step S2 includes the following sub-steps:
s2-1, sampling the received signals for N times, wherein the number of sampling points in each sampling is M, and obtaining N sample sets;
s2-2, according to a formula:
Figure 945524DEST_PATH_IMAGE001
through the first stepiThe average value of the samples in the sample set replaces the expected value, and the average value is obtainediGeometric power corresponding to each sample set
Figure 780887DEST_PATH_IMAGE002
(ii) a Wherein
Figure 277727DEST_PATH_IMAGE003
Denotes the firstiIn a sample setjA sample is obtained;
s2-3, according to a formula:
Figure 928152DEST_PATH_IMAGE004
constructing feature vectors
Figure 101513DEST_PATH_IMAGE005
。
Further, the specific method of step S3 includes the following substeps:
s3-1, according to a formula:
Figure 361855DEST_PATH_IMAGE006
constructing hyperplane equation of supervised learning model to obtain linear kernel
Figure 396807DEST_PATH_IMAGE007
(ii) a Wherein
Figure 901738DEST_PATH_IMAGE008
Is a weight vector;
Figure 262312DEST_PATH_IMAGE009
is a deviation vector;
Figure 367540DEST_PATH_IMAGE010
represents a transpose of a matrix;
Figure 940604DEST_PATH_IMAGE008
and
Figure 768883DEST_PATH_IMAGE009
all are parameters to be trained of the supervised learning model; x is the input of the supervised learning model;
s3-2, initialization
Figure 785511DEST_PATH_IMAGE008
And
Figure 128768DEST_PATH_IMAGE009
；
s3-3, taking the feature vector as the input of a supervised learning model, and acquiring the value of a linear kernel corresponding to the feature vector;
s3-4, when the value of the linear kernel corresponding to the feature vector is larger than or equal to 1, outputting a label of a frequency spectrum used by a master user by the supervised learning model; when the value of the linear kernel corresponding to the feature vector is less than or equal to-1, the supervised learning model outputs a label of the unused frequency spectrum of the master user;
s3-5, judging whether the classification success rate of the current supervised learning model reaches an expected value, and if so, taking the current supervised learning model as the supervised learning model after training; otherwise, entering step S3-6;
s3-6, constructing a loss function, calculating a loss value through a real label of the feature vector and an output label of the supervised learning model, and reversely propagating and updating
Figure 771102DEST_PATH_IMAGE008
And
Figure 250625DEST_PATH_IMAGE009
and returning to the step S3-3.
Further, the specific method for acquiring the geometric power of the signal in the actual environment in step S4 is as follows:
the geometric power of the signal in the actual environment is obtained in the same manner as in step S2.
Further, the supervised learning model in step S3 includes an SVM model and a KNN model.
The invention has the beneficial effects that: according to the method, the geometric power of the received signal is obtained to serve as the basis for sensing whether the frequency spectrum is used, and compared with a mode of sensing the frequency spectrum by using energy detection and differential entropy, the method has better sensing accuracy.
Drawings
FIG. 1 is a schematic flow diagram of the process;
FIG. 2 is a performance comparison diagram of SVM spectrum sensing method based on GP, ED and DE;
fig. 3 is a performance comparison diagram of KNN spectrum sensing method based on GP, ED and DE.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate the understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and it will be apparent to those skilled in the art that various changes may be made without departing from the spirit and scope of the invention as defined and defined by the appended claims, and all changes that can be made by the invention using the inventive concept are intended to be protected.
As shown in fig. 1, the method for sensing supervised spectrum based on geometric power includes the following steps:
s1, constructing a noise model with generalized Gaussian distribution, and simulating and receiving signals in a noise environment generated by the noise model;
s2, solving the geometric power of the received signal, and taking the obtained geometric power as a feature vector;
s3, constructing a supervised learning model, and training the supervised learning model through the feature vectors to obtain a trained supervised learning model;
and S4, acquiring and inputting the geometric power of the signal in the actual environment into the trained supervised learning model for spectrum sensing.
In the step S1, the shape parameter of the noise is more than 0 and less than or equal to 2, and the scale parameter of the noise is more than 0. Probability density function of noise model
Figure 936690DEST_PATH_IMAGE011
Comprises the following steps:
Figure 501664DEST_PATH_IMAGE012
wherein
Figure 619792DEST_PATH_IMAGE013
Is a gamma function;
Figure 970133DEST_PATH_IMAGE014
expressed as natural constantseA base exponential function;xis noise;
Figure 312253DEST_PATH_IMAGE015
representing a set of real numbers.
The specific method of step S2 includes the following substeps:
s2-1, sampling the received signals for N times, wherein the number of sampling points in each sampling is M, and obtaining N sample sets;
s2-2, according to a formula:
Figure 364523DEST_PATH_IMAGE001
through the first stepiThe average value of the samples in the sample set replaces the expected value, and the average value is obtainediGeometric power corresponding to each sample set
Figure 535610DEST_PATH_IMAGE002
(ii) a Wherein
Figure 255304DEST_PATH_IMAGE003
Is shown asiIs concentrated in a samplejA sample is obtained;
s2-3, according to a formula:
Figure 33904DEST_PATH_IMAGE004
constructing feature vectors
Figure 515746DEST_PATH_IMAGE005
。
The specific method of step S3 includes the following substeps:
s3-1, according to a formula:
Figure 37994DEST_PATH_IMAGE006
constructing a hyperplane equation of a supervised learning model to obtain a linear kernel
Figure 346616DEST_PATH_IMAGE007
(ii) a Wherein
Figure 545385DEST_PATH_IMAGE008
Is a weight vector;
Figure 572247DEST_PATH_IMAGE009
is a deviation vector;
Figure 367027DEST_PATH_IMAGE010
represents a transpose of a matrix;
Figure 280888DEST_PATH_IMAGE008
and
Figure 401291DEST_PATH_IMAGE009
all are parameters to be trained of the supervised learning model; x is the input of the supervised learning model;
s3-2, initialization
Figure 118711DEST_PATH_IMAGE008
And
Figure 232029DEST_PATH_IMAGE009
；
s3-3, taking the feature vector as the input of a supervised learning model, and acquiring the value of a linear kernel corresponding to the feature vector;
s3-4, when the value of the linear kernel corresponding to the feature vector is larger than or equal to 1, outputting a label of a frequency spectrum used by a master user by the supervised learning model; when the value of the linear kernel corresponding to the feature vector is less than or equal to-1, outputting a label of the unused frequency spectrum of the main user by the supervised learning model;
s3-5, judging whether the classification success rate of the current supervised learning model reaches an expected value or not, and if so, taking the current supervised learning model as the trained supervised learning model; otherwise, entering step S3-6;
s3-6, constructing a loss function, calculating a loss value through a real label of the feature vector and an output label of the supervised learning model, and reversely propagating and updating
Figure 515243DEST_PATH_IMAGE008
And
Figure 72126DEST_PATH_IMAGE009
and returning to the step S3-3.
The specific method for acquiring the geometric power of the signal in the actual environment in step S4 is as follows: the geometric power of the signal in the actual environment is obtained in the same manner as in step S2.
The supervised learning model in the step S3 comprises an SVM model and a KNN model, and the SVM model is preferentially adopted.
In an embodiment of the present invention, an SVM and a KNN are respectively used as a supervised learning model, as shown in fig. 2 and fig. 3 (the abscissa is a signal-to-noise ratio, and the ordinate is a detection accuracy), the method has a large difference in performance between a method using geometric power (gp) as a feature vector and a method using energy statistics (es) and differential entropy (de) as a feature vector, specifically, when the signal-to-noise ratio is lower than-25 dB, the detection accuracy of the method is far higher than that of the method using energy statistics (es) and differential entropy (de), and the method starts at a signal-to-noise ratio of-30 dB, and can achieve a detection accuracy close to 100%, and the detection effect of the method is better than that of the other two methods.

Claims (3)

1. A supervised spectrum sensing method based on geometric power is characterized by comprising the following steps:
s1, constructing a noise model with generalized Gaussian distribution, and simulating and receiving signals in a noise environment generated by the noise model;
s2, solving the geometric power of the received signal, and taking the obtained geometric power as a feature vector;
s3, constructing a supervised learning model, and training the supervised learning model through the feature vectors to obtain a trained supervised learning model;
s4, acquiring and inputting the geometric power of the signal in the actual environment into the trained supervised learning model for spectrum sensing;
the specific method of step S2 includes the following substeps:
s2-1, sampling the received signals for N times, wherein the number of sampling points in each sampling is M, and obtaining N sample sets;
s2-2, according to a formula:
Figure QLYQS_1
replacing the expected value by the average value of the samples in the ith sample set, and obtaining the geometric power corresponding to the ith sample set
Figure QLYQS_2
(ii) a Wherein
Figure QLYQS_3
Representing the jth sample in the ith sample set;
s2-3, according to a formula:
Figure QLYQS_4
constructing feature vectors
Figure QLYQS_5
；
The specific method of step S3 includes the following substeps:
s3-1, according to a formula:
Figure QLYQS_6
constructing hyperplane equation of supervised learning model to obtain linear kernel
Figure QLYQS_7
(ii) a Wherein
Figure QLYQS_8
Is a weight vector;
Figure QLYQS_9
is a deviation vector;
Figure QLYQS_10
represents a transpose of a matrix;
Figure QLYQS_11
and
Figure QLYQS_12
all are parameters to be trained of the supervised learning model; x is the input of the supervised learning model;
s3-2, initialization
Figure QLYQS_13
And
Figure QLYQS_14
；
s3-3, taking the feature vector as the input of a supervised learning model, and acquiring the value of a linear kernel corresponding to the feature vector;
s3-4, when the value of the linear kernel corresponding to the feature vector is larger than or equal to 1, outputting a label of a frequency spectrum used by a master user by the supervised learning model; when the value of the linear kernel corresponding to the feature vector is less than or equal to-1, outputting a label of the unused frequency spectrum of the main user by the supervised learning model;
s3-5, judging whether the classification success rate of the current supervised learning model reaches an expected value or not, and if so, taking the current supervised learning model as the trained supervised learning model; otherwise, entering step S3-6;
s3-6, constructing a loss function, calculating a loss value through a real label of the feature vector and an output label of the supervised learning model, and reversely propagating and updating
Figure QLYQS_15
And
Figure QLYQS_16
returning to the step S3-3;
the specific method for acquiring the geometric power of the signal in the actual environment in step S4 is as follows:
the geometric power of the signal in the actual environment is obtained in the same manner as in step S2.
2. The supervised spectrum sensing method based on geometric power as recited in claim 1, wherein in step S1, a shape parameter of the noise is greater than 0 and less than or equal to 2, and a scale parameter of the noise is greater than 0.
3. The supervised geometric power-based supervised spectrum sensing method according to claim 1, wherein the supervised learning model in step S3 comprises an SVM model and a KNN model.
CN202211470072.3A 2022-11-23 2022-11-23 Supervision spectrum sensing method based on geometric power Active CN115577253B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211470072.3A CN115577253B (en) 2022-11-23 2022-11-23 Supervision spectrum sensing method based on geometric power

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211470072.3A CN115577253B (en) 2022-11-23 2022-11-23 Supervision spectrum sensing method based on geometric power

Publications (2)

Publication Number Publication Date
CN115577253A CN115577253A (en) 2023-01-06
CN115577253B true CN115577253B (en) 2023-02-28

Family

ID=84590741

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211470072.3A Active CN115577253B (en) 2022-11-23 2022-11-23 Supervision spectrum sensing method based on geometric power

Country Status (1)

Country Link
CN (1) CN115577253B (en)

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2004102186A (en) * 2002-09-12 2004-04-02 Matsushita Electric Ind Co Ltd Device and method for sound encoding

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN100346392C (en) * 2002-04-26 2007-10-31 松下电器产业株式会社 Device and method for encoding, device and method for decoding
CN109039503A (en) * 2018-09-07 2018-12-18 广东工业大学 A kind of frequency spectrum sensing method, device, equipment and computer readable storage medium
CN113095162B (en) * 2021-03-24 2023-05-23 杭州电子科技大学 Spectrum sensing method based on semi-supervised deep learning

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2004102186A (en) * 2002-09-12 2004-04-02 Matsushita Electric Ind Co Ltd Device and method for sound encoding

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
Ｌａｐｌａｃｉａｎ 噪声下认知物联网频谱感知算法;刘小蒙等;《计算机应用研究》;20201231;第279-280页 *
利用功率谱极值和几何平均的频谱感知算法;韩仕鹏等;《信号处理》;20181025(第10期);全文 *
基于功率谱几何平均的频谱感知算法;谢起楠等;《杭州电子科技大学学报(自然科学版)》;20200715(第04期);全文 *
基于浊音语音谐波谱子带加权重建的抗噪声说话人识别;曾毓敏等;《东南大学学报(自然科学版)》;20081120(第06期);全文 *

Also Published As

Publication number Publication date
CN115577253A (en) 2023-01-06

Similar Documents

Publication Publication Date Title
CN110364144B (en) Speech recognition model training method and device
Zhang et al. NAS-AMR: Neural architecture search-based automatic modulation recognition for integrated sensing and communication systems
Chen et al. Temporal and spectral feature learning with two-stream convolutional neural networks for appliance recognition in NILM
Zhang et al. A novel automatic modulation classification scheme based on multi-scale networks
CN107703480A (en) Mixed kernel function indoor orientation method based on machine learning
CN113315593A (en) Frequency spectrum sensing algorithm based on FLOM covariance matrix and LSTM neural network
How et al. SOC estimation using deep bidirectional gated recurrent units with tree parzen estimator hyperparameter optimization
CN115577253B (en) Supervision spectrum sensing method based on geometric power
Geng et al. Spectrum sensing for cognitive radio based on feature extraction and deep learning
CN111669820B (en) Density peak value abnormity detection method and intelligent passive indoor positioning method
CN114337883B (en) CNN collaborative spectrum sensing method and system for covariance matrix Cholesky decomposition
CN115130620B (en) Power equipment power utilization mode identification model generation method and device
Zhuang et al. Siegel distance-based fusion strategy and differential evolution algorithm for cooperative spectrum sensing
CN116090390A (en) FINFET device direct current characteristic prediction method based on deep learning
Wang et al. A geodesic projection-based data fusion scheme for cooperative spectrum sensing
Qin Software reliability prediction model based on PSO and SVM
Wu et al. A novel hybrid particle swarm optimization for feature selection and kernel optimization in support vector regression
Li et al. A spectrum sensing algorithm based on correlation coefficient and K-means
Zhuang et al. Blind spectrum sensing based on the statistical covariance matrix and k-median clustering algorithm
Yang et al. TRAIL: A Three-Step Robust Adversarial Indoor Localization Framework
Zhao et al. Leveraging Topic Model for CSI Based Human Activity Recognition
Huang et al. A method for extracting fingerprint feature of communication satellite signal
Du et al. Device-Free Indoor Localization Based on Multidimensional CSI Features Classification
Chen et al. A Novel Method Based on Random Matrix Theory and Mean Shift Clustering for Spectrum Sensing
Bian et al. LDA enhanced one-bit compressive sensing method for high-throughput mass spectrometry data feature selection

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