CN109104257B - Wireless signal detection method and device - Google Patents

Wireless signal detection method and device Download PDF

Info

Publication number
CN109104257B
CN109104257B CN201810723748.2A CN201810723748A CN109104257B CN 109104257 B CN109104257 B CN 109104257B CN 201810723748 A CN201810723748 A CN 201810723748A CN 109104257 B CN109104257 B CN 109104257B
Authority
CN
China
Prior art keywords
wireless signal
feature
spectrum
frequency band
sample
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
CN201810723748.2A
Other languages
Chinese (zh)
Other versions
CN109104257A (en
Inventor
冯志勇
黄赛
鲁广成
张奇勋
张轶凡
尉志青
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.)
Beijing University of Posts and Telecommunications
Original Assignee
Beijing University of Posts and Telecommunications
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 Beijing University of Posts and Telecommunications filed Critical Beijing University of Posts and Telecommunications
Priority to CN201810723748.2A priority Critical patent/CN109104257B/en
Publication of CN109104257A publication Critical patent/CN109104257A/en
Application granted granted Critical
Publication of CN109104257B publication Critical patent/CN109104257B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/30—Monitoring; Testing of propagation channels
    • H04B17/309—Measuring or estimating channel quality parameters
    • H04B17/318—Received signal strength
    • H04B17/327—Received signal code power [RSCP]
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Electromagnetism (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Quality & Reliability (AREA)
  • Mobile Radio Communication Systems (AREA)
  • Monitoring And Testing Of Transmission In General (AREA)

Abstract

The embodiment of the invention provides a wireless signal detection method and a wireless signal detection device, which relate to the technical field of cognitive radio, wherein the method comprises the following steps: acquiring a wireless signal to be detected; obtaining target frequency bands of spectral features to be extracted, obtaining the types of the spectral features to be extracted on each target frequency band as the characteristic types of each target frequency band, wherein a characteristic selection algorithm based on characteristic sorting is adopted to process samples, and the characteristic types of the target frequency bands and the target frequency bands are determined; extracting the frequency spectrum characteristics of the wireless signal to be detected, which belong to the characteristic category of the target frequency band on the target frequency band, aiming at each target frequency band; and detecting whether the wireless signal to be detected belongs to the target type or not according to the extracted frequency spectrum characteristics. By applying the embodiment of the invention, the wireless signal of the target type can be identified with higher accuracy than that of the prior art in the application scene with the interfering wireless signal.

Description

Wireless signal detection method and device
Technical Field
The invention relates to the technical field of cognitive radio, in particular to a wireless signal detection method and device.
Background
Along with the development of unmanned aerial vehicle trade, its application is more and more extensive, and unmanned aerial vehicle signal detection receives more and more attention. Only after the signal detection work of the unmanned aerial vehicle is done, the follow-up work of returning, capturing and the like of the unmanned aerial vehicle can be better carried out.
Conventional wireless signal detection methods include waveform detection and energy detection. The following energy detection methods are generally adopted in the prior art to detect unmanned aerial vehicle signals: the method comprises the following steps of utilizing a window function which can cover at least 3 frequency points, translating from low frequency to high frequency on a power spectrum of a wireless signal to be detected, calculating a signal gradient corresponding to the frequency point covered by a primary window function when translating one frequency point, judging that the wireless signal to be detected is a suspected unmanned aerial vehicle signal if the power spectrum of the wireless signal to be detected accords with the following three conditions, and judging that the wireless signal to be detected is not an unmanned aerial vehicle signal if the power spectrum of the wireless signal to be detected accords with the following three conditions:
in the first case: when the window gradually enters a change edge area, the signal gradient is changed from small to large;
in the second case: when the whole window is completely in the change edge region, the signal gradient obtains the maximum value and certain stationarity is kept;
in the third case: when the window leaves the change edge area, the signal gradient is changed from large to small.
The method requires that the frequency band of the wireless signal to be detected is relatively pure and has no other interference signals. While the 2.4GHz band where most of the signals of the unmanned aerial vehicle are located is a common ISM (Industrial Scientific Medical) band, and various wireless signals such as Wi-Fi signals and bluetooth signals exist in the band. The existence of the interference signal influences the accuracy of the detection of the method, so that the accuracy of the detection result is low.
Disclosure of Invention
The embodiment of the invention aims to provide a wireless signal detection method and a wireless signal detection device, so that unmanned aerial vehicle signals can be identified with higher accuracy than that of the prior art in an application scene with interference wireless signals.
The specific technical scheme is as follows:
the embodiment of the invention provides a wireless signal detection method, which comprises the following steps:
acquiring a wireless signal to be detected;
acquiring target frequency bands of spectral features to be extracted, and acquiring the types of the spectral features to be extracted on each target frequency band as the feature types of each target frequency band;
extracting the frequency spectrum characteristics of the wireless signal to be detected, which belong to the characteristic category of the target frequency band on the target frequency band, aiming at each target frequency band;
detecting whether the wireless signal to be detected belongs to a target type according to the extracted frequency spectrum characteristics;
the target frequency band and the characteristic types of the target frequency band are determined in the following modes:
obtaining a first sample wireless signal, wherein the first sample wireless signal comprises: a wireless signal belonging to a target type and a wireless signal not belonging to the target type;
extracting the spectrum characteristic of the first sample wireless signal as a first sample spectrum characteristic;
obtaining a correlation coefficient corresponding to each cluster of spectrum features in the first sample spectrum feature, wherein the spectrum features contained in each cluster of spectrum features are the same in frequency band and the same in type, and the correlation coefficient corresponding to each cluster of spectrum features is used for characterizing: the frequency band of the frequency spectrum features is in the cluster of frequency spectrum features, and the type of the frequency spectrum features is the confidence coefficient that the wireless signals of the type to which the cluster of frequency spectrum features belongs belong to the target type;
preferentially selecting the correlation coefficient with high represented confidence from the obtained correlation coefficients;
and aiming at each selected correlation coefficient, taking the frequency band of the spectrum feature in the spectrum feature cluster corresponding to the correlation coefficient as a target frequency band, and taking the type of the spectrum feature as the feature type of the target frequency band.
In one implementation manner of the present invention, the extracting a spectrum feature of the first sample wireless signal as a first sample spectrum feature includes:
respectively extracting the power spectral density of each first sample wireless signal on a preset frequency point in a preset frequency band;
respectively calculating the spectrum characteristics of each first sample wireless signal on each sub-band of the preset frequency band based on the extracted power spectral density to serve as the spectrum characteristics of the sub-band;
aiming at each sub-band spectrum feature, obtaining a two-dimensional vector formed by the sub-band spectrum feature and a zero mark corresponding to the sub-band spectrum feature, wherein the zero mark corresponding to each sub-band spectrum feature is as follows: a flag, expressed in the form of a zero, indicating whether the first sample radio signal whose spectral characteristics include those of the sub-band belongs to the target type;
and aggregating the obtained two-dimensional vectors to obtain the first sample spectrum characteristic.
In an implementation manner of the present invention, the preferentially selecting a correlation coefficient with a high characterized confidence degree from the obtained correlation coefficients includes:
performing traversal selection on each obtained correlation coefficient according to the following mode, and determining the correlation coefficient with high represented confidence coefficient in the obtained correlation coefficients:
selecting the largest unselected correlation coefficient from the obtained correlation coefficients as the current correlation coefficient;
adding the spectrum feature cluster corresponding to the current correlation coefficient to a traversal feature set;
taking the labeling information corresponding to each spectrum feature in the traversal feature set and the traversal feature set as the input of a first quadratic logistic regression model, selecting a model parameter which enables the accuracy of the detection result of the first quadratic logistic regression model to be highest based on a ten-fold cross-validation method, and finishing the training of the first quadratic logistic regression model, wherein the labeling information corresponding to each spectrum feature represents: whether the first sample wireless signal with the spectrum characteristic is a target type wireless signal or not, wherein the detection result of the first two-term logistic regression model is used for representing that: the frequency spectrum characteristic is whether the wireless signal of each frequency spectrum characteristic in the traversal characteristic set belongs to a target type;
judging whether the accuracy of the detection result corresponding to the selected model parameter is greater than the accuracy of the historical detection result;
if not, deleting the spectrum feature cluster corresponding to the current correlation coefficient from the traversal feature set;
and if so, determining the current correlation coefficient as one of the obtained correlation coefficients with high represented confidence.
In an implementation manner of the present invention, the feature type of the target frequency band includes at least one of the following feature types:
mean of power spectral density;
variance of power spectral density;
differential characterization of power spectral density.
In an implementation manner of the present invention, the detecting whether the wireless signal to be detected belongs to a target type according to the extracted spectral feature includes:
inputting the extracted spectral features into a classifier obtained by pre-training to obtain a detection result of whether the wireless signal to be detected belongs to a target type;
wherein the two classifiers are obtained by training in the following way:
obtaining a second sampled wireless signal, wherein the second sampled wireless signal comprises: a wireless signal belonging to a target type and a wireless signal not belonging to the target type;
extracting the frequency spectrum characteristics of the second sample wireless signals on the target frequency band, belonging to the characteristic type of the target frequency band, as second sample frequency spectrum characteristics aiming at each target frequency band;
for each second sample wireless signal, obtaining a feature vector formed by the spectrum features belonging to the second sample wireless signal in the second sample spectrum features;
and constructing a second-term logistic regression model by taking the obtained feature vectors and a zero mark corresponding to each feature vector as known quantities and taking the weight parameter vectors and the bias parameters as unknown quantities, wherein the zero mark corresponding to each feature vector is as follows: a flag, expressed in the form of a zero, indicating whether the second sample wireless signal corresponding to each feature vector belongs to the target type;
obtaining the second logistic regression model to enable the maximum likelihood algorithm to obtain the value of the weight parameter and the value of the bias parameter when the maximum value is obtained;
and determining a second-term logistic regression model when the values of the weight parameter vector and the bias parameter are respectively obtained values as the two classifiers.
In an implementation manner of the present invention, the constructing a second logistic regression model by using the obtained feature vectors and a zero flag corresponding to each feature vector as known quantities and using the weight parameter vector and the bias parameter as unknown quantities includes:
establishing the second term logistic regression model by adopting the following formula:
Figure GDA0002528374350000041
wherein x isiIs the i-th feature vector, yiA zero mark corresponding to the ith feature vector, ω is the weight parameter vector, and b is the bias parameter.
In an implementation manner of the present invention, the obtaining of the second logistic regression model enables the value of the weight parameter and the value of the bias parameter when the maximum likelihood algorithm obtains the maximum value, including:
obtaining the second binomial logistic regression model by adopting a gradient descent method to calculate the maximum value of a preset log-likelihood function, so that the weight parameter value and the bias parameter value are obtained when the maximum value is obtained by a maximum likelihood algorithm;
wherein the preset log-likelihood function is:
Figure GDA0002528374350000051
wherein L (ω) is a log-likelihood function, xiIs the i-th feature vector, yiA zero mark corresponding to the ith eigenvector, N is the number of eigenvectors, omega is the weight parameter vector, and b is the bias parameter.
In an implementation manner of the present invention, the determining, as the second classifier, a second logistic regression model when the values of the weight parameter vector and the bias parameter are respectively obtained values includes:
determining a classifier represented by the following expression as the two classifiers:
Figure GDA0002528374350000052
wherein，
Figure GDA0002528374350000053
The value of the weight parameter vector is taken as the value,
Figure GDA0002528374350000054
taking the value of the bias parameter, wherein x is a vector which is input into the two classifiers and is formed by spectrum features, y is a zero mark output by the two classifiers, and the zero mark output by the two classifiers is: and a flag, expressed in the form of a zero, indicating whether the wireless signal represented by the vector input to the two classifiers belongs to the target type.
An embodiment of the present invention further provides a wireless signal detection apparatus, including:
the signal acquisition module is used for acquiring a wireless signal to be detected;
the frequency band obtaining module is used for obtaining a target frequency band of the spectral features to be extracted and obtaining the type of the spectral features to be extracted on each target frequency band as the characteristic type of each target frequency band;
the characteristic extraction module is used for extracting the frequency spectrum characteristics of the wireless signal to be detected, which belong to the characteristic category of the target frequency band on the target frequency band, aiming at each target frequency band;
the signal detection module is used for detecting whether the wireless signal to be detected belongs to a target type according to the extracted frequency spectrum characteristics;
wherein, the frequency band obtaining module includes:
a sample signal acquisition submodule for obtaining a first sample wireless signal, wherein the first sample wireless signal comprises: a wireless signal belonging to a target type and a wireless signal not belonging to the target type;
the sample characteristic extraction submodule is used for extracting the frequency spectrum characteristic of the first sample wireless signal as a first sample frequency spectrum characteristic;
a correlation coefficient obtaining sub-module, configured to obtain a correlation coefficient corresponding to each cluster of spectral features in the first sample spectral feature, where the spectral features included in each cluster have the same frequency band and the same type of spectral features, and the correlation coefficient corresponding to each cluster of spectral features is used to characterize: the frequency band of the frequency spectrum features is in the cluster of frequency spectrum features, and the type of the frequency spectrum features is the confidence coefficient that the wireless signals of the type to which the cluster of frequency spectrum features belongs belong to the target type;
the correlation coefficient optimization submodule is used for preferentially selecting the correlation coefficient with high represented confidence from the obtained correlation coefficients;
and the target frequency band determining submodule is used for taking the frequency band of the frequency spectrum features in the frequency spectrum feature cluster corresponding to the correlation coefficient as the target frequency band and taking the types of the frequency spectrum features as the feature types of the target frequency band aiming at each selected correlation coefficient.
The embodiment of the invention also provides electronic equipment which comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory complete mutual communication through the communication bus;
a memory for storing a computer program;
and a processor for implementing any of the steps of the wireless signal detection method described above when executing the program stored in the memory.
In yet another aspect of the present invention, the present invention further provides a computer-readable storage medium, which stores instructions that, when executed on a computer, cause the computer to perform any of the steps of the wireless signal detection method described above.
In yet another aspect of the present invention, the present invention also provides a computer program product containing instructions, which when run on a computer, causes the computer to execute any of the above-mentioned wireless signal detection methods.
As can be seen from the above, in the scheme provided in the embodiment of the present invention, during training, a feature selection algorithm based on feature sorting is used for the wireless signals belonging to the target type and the wireless signals not belonging to the target type, so as to obtain reliable spectrum features for detecting the target type wireless signals through screening. During detection, only reliable frequency spectrum features are extracted for detection based on training results, and the influence of interference signals on detection can be eliminated, so that the wireless signals of the target type can be identified with higher accuracy than that of the prior art in an application scene with interference wireless signals. Of course, not all of the advantages described above need to be achieved at the same time in the practice of any one product or method of the invention.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a flowchart of a wireless signal detection method according to an embodiment of the present invention;
fig. 2 is a flowchart of an information determining method according to an embodiment of the present invention;
FIG. 3 is a flow chart of a method for selecting correlation coefficients according to an embodiment of the present invention;
FIG. 4 is a flowchart of a classifier training method according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a wireless signal detection apparatus according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of an obtaining module according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The embodiment of the present invention provides a wireless signal detection method, and the following describes in detail the wireless signal detection method provided by the embodiment of the present invention through a specific embodiment.
Referring to fig. 1, fig. 1 is a flowchart of a wireless signal detection method according to an embodiment of the present invention, including the following steps:
and S101, acquiring a wireless signal to be detected.
In the embodiment of the invention, a wireless signal data acquisition platform constructed by equipment such as a frequency spectrograph and an omnidirectional antenna can be used for acquiring the wireless signal to be detected, and the wireless signal to be detected acquired by the equipment can also be acquired in a wired mode.
Step S102, obtaining target frequency bands of the spectral features to be extracted, and obtaining the types of the spectral features to be extracted on each target frequency band as the feature types of each target frequency band.
Specifically, the samples may be processed by a feature selection algorithm based on feature sorting to determine the target frequency band and the feature type of the target frequency band.
In the embodiment of the invention, in order to eliminate the influence of interference signals, only a plurality of specific frequency bands are selected for detection on the frequency bands possibly related to the detection of the target type wireless signals, and the frequency bands are taken as the target frequency bands. And on each target frequency band, only one or more spectral features with different spectral feature types are selected for detection. The categories of spectral features may include power features, power difference features, energy features, waveform variation features, and the like.
Step S103, extracting the frequency spectrum characteristics of the wireless signal to be detected on the target frequency band, which belong to the characteristic type of the target frequency band, aiming at each target frequency band.
In order to prevent the detection time consumption and the operation cost from increasing due to the excessive number of the spectral features, in an embodiment of the present invention, when the spectral features of the wireless signal to be detected on one target frequency band are extracted, only one spectral feature of each type of spectral features may be extracted.
And step S104, detecting whether the wireless signal to be detected belongs to the target type or not according to the extracted spectral characteristics.
The target type wireless signal can be an unmanned aerial vehicle signal, and can also be other types of wireless signals.
In the embodiment of the invention, the extracted spectral characteristics can be compared with the manually set threshold value to judge whether the wireless signal to be detected belongs to the target type, and the extracted spectral characteristics can also be input into a classifier obtained by pre-training to obtain the detection result whether the wireless signal to be detected belongs to the target type. Among them, the binary classifier is a device that classifies input data into two classes.
In step S102, a method for processing the sample by using a feature selection algorithm based on feature sorting to determine a target frequency band and a feature type of the target frequency band is described with reference to fig. 2. Fig. 2 provides an information determining method, by which a target frequency band and a feature type of the target frequency band can be determined, including the following steps:
step 201, obtaining a first sample wireless signal.
The first sample wireless signal includes: a wireless signal belonging to a target type and a wireless signal not belonging to a target type.
When wireless signals which do not belong to the target type are collected, the application scene of actual detection can be simulated, and wireless environment signals are collected. When the wireless signals belonging to the target type are collected, the target type wireless signals can be used for communication in the original channel environment without the target type wireless model, and then the wireless signals are collected in the environment. The specific acquisition method is the same as that for acquiring the wireless signal to be detected in step S101.
Step 202, extracting the spectrum feature of the first sample wireless signal as the first sample spectrum feature.
For the convenience of training, each of the first sample spectral features may be labeled and extracted from a wireless signal belonging to a target type or a wireless signal not belonging to the target type. The signature is also taken as part of the spectral signature of the first sample.
And step 203, obtaining a correlation coefficient corresponding to each cluster of spectral features in the first sample spectral features.
Each frequency spectrum characteristic in the same frequency spectrum characteristic cluster is the same in frequency band and the type of the frequency spectrum characteristic is the same. Each spectral feature in the same cluster of spectral features is extracted from a different one of the first sample wireless signals.
And the correlation coefficient corresponding to the spectrum feature cluster is the spectrum feature in the spectrum feature cluster and a zero-labeled correlation coefficient for extracting whether the wireless signal of the spectrum feature belongs to the target type. And the correlation coefficient corresponding to the spectrum feature cluster is used for characterizing: the frequency band of the frequency spectrum characteristic is in the cluster of frequency spectrum characteristics, and the type of the frequency spectrum characteristic is the confidence coefficient that the wireless signal of the type to which the cluster of frequency spectrum characteristics belongs to the target type.
In one embodiment of the invention, Pearson correlation coefficients may be used to characterize: the frequency band of the frequency spectrum characteristic is in the cluster of frequency spectrum characteristics, and the type of the frequency spectrum characteristic is the confidence coefficient that the wireless signal of the type to which the cluster of frequency spectrum characteristics belongs to the target type.
One spectral feature cluster may be represented by the following set:
Uk＝{(x1,k,y1),(x2,k,y2),…,(xi,k,yi),…,(xn,k,yn)}
wherein, UkFor the kth spectral feature cluster, xi,kIs the k-th spectral characteristic, y, of the i-th wireless signal in the first sample wireless signaliN is the number of wireless signals in the first sample wireless signal, in order to indicate the mark whether the ith wireless signal in the first sample wireless signal belongs to the target type in a zero form.
If the ith wireless signal is of the target type, then yi1, if the ith wireless signal does not belong to the target type, then yi＝0。
The pearson correlation coefficient corresponding to each cluster of the first sample spectrum feature can be calculated by the following formula:
Figure GDA0002528374350000101
where ρ iskPearson's correlation coefficient, x, for the kth spectral feature clusterkIs the spectral feature of the kth spectral feature cluster, y is a flag indicating whether the wireless signal in the first sample wireless signal belongs to the target type in the form of a zero,
Figure GDA0002528374350000102
is xkThe average value of (a) of (b),
Figure GDA0002528374350000103
is the average value of the values of y,
Figure GDA0002528374350000104
is xkStandard deviation of (a)yIs the standard deviation of y.
And step 204, preferentially selecting the correlation coefficient with high represented confidence from the obtained correlation coefficients.
In the embodiment of the present invention, a preset number of correlation coefficients with the highest correlation coefficient may be selected, or all correlation coefficients with correlation coefficient values larger than a preset threshold may be selected.
If the correlation coefficient is not selected, all the frequency spectrum features are used for training, the time cost and the operation cost of training can be increased, the influence of interference signals in a part of frequency spectrum features is large, the correlation coefficient is reduced, the frequency spectrum features are removed, and the detection accuracy is improved.
Step 205, regarding each selected correlation coefficient, taking the frequency band of the spectrum feature in the spectrum feature cluster corresponding to the correlation coefficient as a target frequency band, and taking the type of the spectrum feature as the feature type of the target frequency band.
As can be seen from the above, in the scheme provided in the embodiment of the present invention, during training, a feature selection algorithm based on feature sorting is used for the wireless signals belonging to the target type and the wireless signals not belonging to the target type, so as to obtain reliable spectrum features for detecting the target type wireless signals through screening. During detection, only reliable frequency spectrum features are extracted for detection based on training results, and the influence of interference signals on detection can be eliminated, so that the wireless signals of the target type can be identified with higher accuracy than that of the prior art in an application scene with interference wireless signals.
In an implementation manner of the present invention, the extracting of the spectrum feature of the first sample wireless signal in the step 202 may be implemented as the following steps a 1-a 4:
step A1: and respectively extracting the power spectral density of each first sample wireless signal at a preset frequency point in a preset frequency band.
In the embodiment of the invention, the preset frequency band is all frequency bands possibly related to the detection target type wireless signal, the general range is large, and the frequency bands possibly useful for detecting the target type wireless signal are all included.
The power spectral density is the power carried per unit frequency wave and is used to indicate how the power of a signal is distributed with frequency. The detection of the wireless signal based on the present implementation belongs to power feature detection using the calculation result based on the power spectral density as the spectral feature.
Step A2: and respectively calculating the spectrum characteristics of each first sample wireless signal on each sub-band of the preset frequency band based on the extracted power spectral density as the spectrum characteristics of the sub-band.
In the embodiment of the invention, the preset frequency band can be uniformly divided to obtain a plurality of sub-frequency bands with equal width. The frequency spectrum characteristics of the sub-band can be calculated by a statistical formula based on the power spectrum density of a plurality of preset frequency points in the sub-band. Different descriptive statistics of the power spectral densities of a plurality of preset frequency points, namely different kinds of sub-frequency band spectrum characteristics. For each sub-band there is only one sub-band spectral feature of each class.
Step A3: and aiming at each sub-band spectrum characteristic, obtaining a two-dimensional vector formed by the sub-band spectrum characteristic and a zero mark corresponding to the sub-band spectrum characteristic.
A zero corresponding to each sub-band spectrum feature is marked as: a flag, expressed in the form of a zero, indicating whether the first sample radio signal whose spectral characteristics include spectral characteristics of the sub-band belongs to the target type. When the value of the zero marker is one, the first sample wireless signal of which the spectrum characteristic contains the sub-band spectrum characteristic belongs to the target type, and when the value of the zero marker is zero, the first sample wireless signal of which the spectrum characteristic contains the sub-band spectrum characteristic does not belong to the target type.
In the embodiment of the present invention, the two-dimensional vector may be represented as: (x)i,k,yi) Wherein x isi,kIs the k-th spectral characteristic, y, of the i-th wireless signal in the first sample wireless signaliIs a flag indicating whether the ith wireless signal in the first sample wireless signal belongs to the target type in the form of a zero.
Step A4: and aggregating the obtained two-dimensional vectors to obtain a first sample spectrum characteristic.
In this embodiment of the present invention, the first sample spectrum feature may be represented by a two-dimensional vector set as follows:
D＝{(x1,1,y1),(x1,2,y1),…,(x1,3s,y1),(x2,1,y2),…,(xi,k,yi),…,(xn,3s,yn)}
wherein D is the first sample spectral feature.
In the implementation mode, when the training data is extracted from the samples, whether the first sample wireless signal belongs to the target type or not is combined with the frequency spectrum characteristic data to form a two-dimensional vector, and the two-dimensional vector can be directly used for calculating the correlation coefficient during the subsequent sample processing of the characteristic selection algorithm based on the characteristic sorting, so that the data processing in the training is facilitated.
In an implementation manner of the present invention, the feature type of the target frequency band in step S102 includes at least one of the following three feature types: mean of power spectral density; variance of power spectral density; differential characterization of power spectral density.
Then, the step a2 of calculating the descriptive statistic by using the power spectral density of the plurality of preset frequency points in the sub-band may include:
calculating the mean value of the power spectral density over each sub-band using the following formula:
Figure GDA0002528374350000121
wherein,
Figure GDA0002528374350000122
is the mean value of the power spectral density of the mth sub-band, s is the number of the sub-bands obtained by dividing the preset frequency band, a is the number of the preset frequency points on the mth sub-band, P isjIs the power spectral density of the j frequency point on the m frequency sub-band.
The variance of the power spectral density over each sub-band is calculated using the following equation:
Figure GDA0002528374350000123
wherein,
Figure GDA0002528374350000124
is the variance of the power spectral density of the mth frequency sub-band.
Calculating the differential characteristics of the power spectral density at each sub-band using the following formula:
Figure GDA0002528374350000131
wherein, Δ PmIs a differential characterization of the power spectral density of the mth sub-band, Pj+1Is the power spectral density of the j +1 frequency point on the mth frequency sub-band.
The first sample spectrum characteristic in the step a4 can be expressed by the following formula:
D＝{(x1,1,y1),(x1,2,y1),…,(x1,3s,y1),(x2,1,y2),…,(xi,k,yi),…,(xn,3s,yn)}
Figure GDA0002528374350000132
wherein,
Figure GDA0002528374350000133
is the mean value, sigma, of the power spectral density of the ith radio signal in the first sample radio signal in the s-th sub-band2 i,sIs the variance, Δ P, of the power spectral density of the ith radio signal in the first sample radio signal over the s-th sub-bandi,sIs a differential characteristic of the power spectral density of the ith radio signal in the first sample radio signal over the s-th sub-band.
The preset frequency band is uniformly divided into s sub-frequency bands, and each sub-frequency band has the characteristics of the mean value of the power spectral density, the variance of the power spectral density and the difference of the power spectral density, and 3 kinds of frequency spectral characteristics, so that 3s frequency spectral characteristics are provided for the same wireless signal in the first sample wireless signal. So the cluster of spectral features U in step 203kThere are also 3 s.
In this implementation, the spectrum power characteristic and the power difference characteristic of the signal are used to detect and identify the signal of the unmanned aerial vehicle. Compared with the energy characteristics and waveform change characteristics of the signals, the method has the advantages of simple realization and low complexity, can well reflect the energy distribution and jumping conditions of the target type signals in the frequency domain, and has higher detection accuracy. In addition, the characteristic type of the target frequency band is a statistical result obtained by calculating a plurality of power spectral densities in one frequency band through a statistical formula, so that the influence degree of the measurement data of a single frequency point on the detection result is reduced, and the possibility of detection errors caused by measurement errors is reduced.
Fig. 3 is a flowchart of a method for selecting a correlation coefficient according to an embodiment of the present invention, and when the step 204 preferentially selects a correlation coefficient with a high execution degree from the obtained correlation coefficients, each obtained correlation coefficient may be selected in a traversal manner according to the following steps 301 to 308, so as to determine the correlation coefficient with the high execution degree represented in the obtained correlation coefficients.
Specifically, steps 301 to 308 are as follows:
step 301, selecting the largest unselected correlation coefficient from the obtained correlation coefficients as the current correlation coefficient.
In particular, if the values of the plurality of correlation coefficients are equal and maximum, one of the plurality of correlation coefficients may be arbitrarily selected as the current correlation coefficient. If only one correlation coefficient is left unselected among the obtained correlation coefficients, the correlation coefficient may be regarded as the largest correlation coefficient that has not been selected and regarded as the current correlation coefficient.
And step 302, adding the spectrum feature cluster corresponding to the current correlation coefficient to the traversal feature set.
In the embodiment of the present invention, the traversal feature set is a union set of spectrum feature clusters, and may be represented by the following set:
V＝Uα∪Uβ∪…∪Uk
wherein V is the traversal feature set, UαIs the spectral feature cluster with the largest Pearson correlation coefficient, UβFor the spectral feature clusters, U, left in the traversal feature set after the traversal selectionkAnd adding the spectrum feature cluster corresponding to the current correlation coefficient of the traversal feature set for the step.
Before adding the spectral feature cluster with the largest Pearson correlation coefficient to the traversal feature set, the traversal feature set is a null set.
And 303, taking each frequency spectrum feature in the traversal feature set and a zero mark corresponding to the frequency spectrum feature as input of the first quadratic logistic regression model, and selecting a model parameter which enables the detection result accuracy of the first quadratic logistic regression model to be highest based on a cross-over verification method.
The binomial logistic regression model is a binomial discrete selection model based on logical distribution, is the most widely applied binomial model, and the model parameters of the binomial logistic regression model are calculated and valued by using a maximum likelihood estimation method.
The ten-fold cross-validation method is a method for testing the accuracy of an algorithm, and is characterized in that a data set is divided into ten parts, nine parts of the ten parts are taken as training data in turn, and one part of the ten parts is taken as test data to carry out a test.
In the embodiment of the invention, input data of the first quadratic logistic regression model can be divided into ten parts, nine parts of the input data are used as training data to train the first quadratic logistic regression model to obtain values of model parameters, and the remaining part of the input data is used as test data to test the values of the obtained model parameters. The specific test method comprises the following steps: and inputting the frequency spectrum characteristic part of the test data into a first bivariate logistic regression model, wherein the value of the obtained model parameter is used for the model parameter of the first bivariate logistic regression model to obtain the output result of one or zero. And comparing the output one or zero result with the one or zero mark in the corresponding test data, and calculating the frequency of coincidence of the two results to serve as the accuracy of the detection result.
And taking nine parts of the ten data as training data and taking one part of the ten data as test data in turn, and calculating the accuracy of the detection result. And selecting the highest accuracy of the detection results from the ten calculated accuracy of the detection results, and selecting the model parameter which enables the accuracy of the detection results to be highest.
And 304, judging whether the accuracy of the detection result corresponding to the selected model parameter is greater than the accuracy of the historical detection result, if not, entering a step 305, and if so, entering a step 306.
The accuracy of the historical detection result is all the accuracy generated in the process from the beginning of traversal to the current correlation coefficient. If the accuracy of the detection result corresponding to the selected model parameter is greater than the accuracy of the historical detection result, the accuracy is the highest accuracy in the process from the beginning of traversal to the current correlation coefficient.
Particularly, when the step is executed for the first time, only the spectral feature cluster with the largest Pearson correlation coefficient in the feature set is traversed, and the accuracy of the previous historical detection result is not high. At this time, the accuracy of the historical detection result may be regarded as zero, so that when this step is executed for the first time, it is determined that the accuracy is greater than the accuracy of the historical detection result, and the process proceeds to step 306.
Step 305, deleting the spectrum feature cluster corresponding to the current correlation coefficient from the traversal feature set, and entering step 307.
Step 306, determining that the frequency band of the spectrum feature in the spectrum feature cluster corresponding to the current correlation coefficient is a target frequency band, and the type of the spectrum feature is a feature type of the target frequency band.
In the embodiment of the invention, the determination of a target frequency band and the determination of the type of the frequency spectrum characteristic as a characteristic type of the target frequency band can be realized by recording the serial number of the current correlation coefficient, and if the current correlation coefficient is rhokThen the positive integer k is recorded.
Step 307, judging whether all the correlation coefficients are selected, if not, entering step 301, and if so, entering step 308.
And 308, determining the target frequency band and the characteristic type of the target frequency band.
In the embodiment of the present invention, each current correlation coefficient number may be determined through the recording step 306, so as to determine the target frequency band and the feature type of the target frequency band. The value of the current correlation coefficient number k comes from:
Figure GDA0002528374350000161
and aiming at each recorded numerical value of the serial number, finding out a corresponding spectrum characteristic, wherein the frequency band corresponding to the spectrum characteristic is a target frequency band, and the type of the spectrum characteristic is a characteristic type of the target frequency band.
In the implementation mode, the frequency spectrum feature clusters are sequentially added into the traversal training set according to the correlation coefficients from high to low, the detection accuracy after the addition is verified, the frequency spectrum feature clusters with the improved accuracy after the addition are reserved, the frequency spectrum feature clusters which cannot be improved in accuracy are removed, and the correlation coefficients corresponding to the reserved frequency spectrum feature clusters are the correlation coefficients with high represented confidence degrees which are selected preferentially. The spectrum characteristic clusters with high correlation coefficients are combined together, and the detection accuracy is not necessarily high. According to the implementation mode, the selection method of the correlation coefficient is optimized, the correlation coefficient with high correlation coefficient is not simply selected, but the spectrum feature clusters which can improve the detection accuracy by being added into the traversal training set are found through the effect of combining the spectrum features together, so that the correlation coefficient with high represented confidence coefficient is optimized.
Specifically, in the step S104, when detecting whether the wireless signal to be detected belongs to the target type according to the extracted spectral feature, the extracted spectral feature may be input into a pre-trained binary classifier to obtain a detection result of whether the wireless signal to be detected belongs to the target type.
The second classifier can be trained in the following manner as shown in fig. 4.
Fig. 4 is a flowchart of a classifier training method according to an embodiment of the present invention, which specifically includes the following steps:
step 401, a second sample wireless signal is obtained.
The second sample wireless signal includes: a wireless signal belonging to a target type and a wireless signal not belonging to a target type.
In the embodiment of the present invention, the second sample wireless signal may be the same as the first sample signal, or may be different from the first sample signal.
Step 402, extracting, for each target frequency band, a spectrum feature of the second sample wireless signal on the target frequency band, which belongs to the feature type of the target frequency band, as a second sample spectrum feature.
Step 403, for each second sample wireless signal, obtaining a feature vector formed by the spectrum features belonging to the second sample wireless signal in the second sample spectrum features.
In the embodiment of the present invention, the obtained feature vector may be represented as follows:
Figure GDA0002528374350000171
wherein x isiIs the ith obtained feature vector.
If N is the number of wireless signals in the second sample wireless signal, then a total of N eigenvectors can be obtained in this step.
And step 404, constructing a second-term logistic regression model.
In the embodiment of the present invention, the obtained feature vectors and a zero flag corresponding to each feature vector may be used as known quantities, the weight parameter vectors and the bias parameters are used as unknown quantities, and a second term logistic regression model is established by using the following formula:
Figure GDA0002528374350000172
wherein x isiIs the i-th feature vector, yiA zero mark corresponding to the ith eigenvector, omega is the weight parameter vector, and b is the bias parameter.
Step 405, obtaining a second-term logistic regression model to enable the maximum likelihood algorithm to obtain the value of the weight parameter and the value of the bias parameter when the maximum value is obtained.
In the embodiment of the invention, the second logistic regression model can be obtained by adopting a gradient descent method to calculate the maximum value of the preset log-likelihood function, so that the maximum likelihood algorithm obtains the value of the weight parameter and the value of the bias parameter when the maximum value is obtained.
Wherein the preset log-likelihood function is:
Figure GDA0002528374350000173
wherein L (ω) is a log-likelihood function, xiIs the i-th feature vector, yiA zero mark corresponding to the ith eigenvector, N is the number of eigenvectors, omega is the weight parameter vector, and b is the bias parameter.
And step 406, determining the second-term logistic regression model as the second classifier when the values of the weight parameter vector and the bias parameter are respectively the obtained values.
In the embodiment of the present invention, the classifier represented by the following expression may be determined as a binary classifier:
Figure GDA0002528374350000181
wherein,
Figure GDA0002528374350000182
the value of the weight parameter vector is taken as the value,
Figure GDA0002528374350000183
for the value of the bias parameter, x is a vector formed by spectrum features input into the two classifiers, y is a one-zero mark output by the two classifiers, and the one-zero mark output by the two classifiers is: and whether the wireless signal represented by the vector input into the two classifiers belongs to the mark of the target type or not is represented in a form of one zero.
And x is a vector formed by the spectral features input into the two classifiers, and each element of the vector is the spectral feature of the feature type of the wireless signal belonging to each target frequency band on the target frequency band, which is extracted for each target frequency band.
In order to detect whether the wireless signal to be detected belongs to the target type according to the extracted spectral features, the embodiment of the invention provides a training method of a two-classifier, which can input the extracted spectral features into the two-classifier obtained by pre-training to obtain a detection result of whether the wireless signal to be detected belongs to the target type. When the training method of the two-classifier provided by the embodiment of the invention is applied to detecting the wireless signals, the two-term logistic regression model is trained by adopting the wireless signals belonging to the target type and the wireless signals not belonging to the target type to obtain the two-classifier. The spectrum characteristics are input into the classifier for detection, and the detection accuracy is higher than that of the detection which is directly judged by using a manually set threshold value.
Based on the same inventive concept, according to the wireless signal detection method provided in the above embodiment of the present invention, correspondingly, an embodiment of the present invention further provides a wireless signal detection apparatus, a schematic structural diagram of which is shown in fig. 5, and specifically includes:
a signal obtaining module 501, configured to obtain a wireless signal to be detected;
a frequency band obtaining module 502, configured to obtain a target frequency band of a spectral feature to be extracted, and obtain a type of the spectral feature to be extracted on each target frequency band, as a feature type of each target frequency band;
a feature extraction module 503, configured to extract, for each target frequency band, a frequency spectrum feature of the to-be-detected wireless signal, on the target frequency band, that belongs to the feature category of the target frequency band;
a signal detection module 504, configured to detect whether the wireless signal to be detected belongs to a target type according to the extracted spectral feature;
the frequency band obtaining module 502, as shown in fig. 6, includes:
a sample signal obtaining sub-module 601, configured to obtain a first sample wireless signal, where the first sample wireless signal includes: a wireless signal belonging to a target type and a wireless signal not belonging to the target type;
a sample feature extraction sub-module 602, configured to extract a spectral feature of the first sample wireless signal as a first sample spectral feature;
a correlation coefficient obtaining sub-module 603, configured to obtain a correlation coefficient corresponding to each cluster of spectral features in the first sample spectral feature, where the spectral features included in each cluster are the same in frequency band and the types of the spectral features are the same, and the correlation coefficient corresponding to each cluster of spectral features is used to characterize: the frequency band of the frequency spectrum features is in the cluster of frequency spectrum features, and the type of the frequency spectrum features is the confidence coefficient that the wireless signals of the type to which the cluster of frequency spectrum features belongs belong to the target type;
a correlation coefficient preference submodule 604 for preferentially selecting a correlation coefficient with a high characterized confidence from the obtained correlation coefficients;
and a target frequency band determining submodule 605, configured to, for each selected correlation coefficient, use a frequency band in which the spectrum feature in the spectrum feature cluster corresponding to the correlation number is located as a target frequency band, and use a type of the spectrum feature as a feature type in the target frequency band.
The wireless signal detection device provided by the embodiment of the invention is used for detecting wireless signals, and during training, the characteristic selection algorithm based on characteristic sorting is adopted for wireless signals belonging to the target type and wireless signals not belonging to the target type, so that reliable frequency spectrum characteristics for the wireless signals of the detected target type are obtained by screening. During detection, only reliable frequency spectrum features are extracted for detection based on training results, and the influence of interference signals on detection can be eliminated, so that the wireless signals of the target type can be identified with higher accuracy than that of the prior art in an application scene with interference wireless signals.
Based on the same inventive concept, according to the wireless signal detection method provided by the above embodiment of the present invention, correspondingly, the embodiment of the present invention further provides an electronic device, as shown in fig. 7, comprising a processor 701, a communication interface 702, a memory 703 and a communication bus 704, wherein the processor 701, the communication interface 702 and the memory 703 complete mutual communication through the communication bus 704,
a memory 703 for storing a computer program;
the processor 701 is configured to implement the steps of any of the wireless signal detection methods in the above embodiments when executing the program stored in the memory 703.
The communication bus mentioned in the electronic device may be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The communication bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is shown, but this does not mean that there is only one bus or one type of bus.
The communication interface is used for communication between the electronic equipment and other equipment.
The Memory may include a Random Access Memory (RAM) or a Non-Volatile Memory (NVM), such as at least one disk Memory. Optionally, the memory may also be at least one memory device located remotely from the processor.
The Processor may be a general-purpose Processor, including a Central Processing Unit (CPU), a Network Processor (NP), and the like; but may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic device, discrete hardware component.
The wireless signal detection electronic equipment provided by the embodiment of the invention is used for detecting the wireless signals, and during training, the characteristic selection algorithm based on characteristic sorting is adopted for the wireless signals belonging to the target type and the wireless signals not belonging to the target type, so that the reliable frequency spectrum characteristics of the wireless signals for detecting the target type are obtained by screening. During detection, only reliable frequency spectrum features are extracted for detection based on training results, and the influence of interference signals on detection can be eliminated, so that the wireless signals of the target type can be identified with higher accuracy than that of the prior art in an application scene with interference wireless signals.
In yet another embodiment of the present invention, a computer-readable storage medium is further provided, which stores instructions that, when executed on a computer, cause the computer to perform the steps of any of the wireless signal detection methods in the above embodiments.
In yet another embodiment, a computer program product containing instructions is provided, which when run on a computer, causes the computer to perform any of the above-described wireless signal detection methods.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
All the embodiments in the present specification are described in a related manner, and the same and similar parts among the embodiments may be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the apparatus, the electronic device, the computer-readable storage medium, and the computer program product embodiments, since they are substantially similar to the method embodiments, the description is relatively simple, and in relation to them, reference may be made to the partial description of the method embodiments.
The above description is only for the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention shall fall within the protection scope of the present invention.

Claims (10)

1. A method for wireless signal detection, comprising:
acquiring a wireless signal to be detected;
acquiring target frequency bands of spectral features to be extracted, and acquiring the types of the spectral features to be extracted on each target frequency band as the feature types of each target frequency band;
extracting the frequency spectrum characteristics of the wireless signal to be detected, which belong to the characteristic category of the target frequency band on the target frequency band, aiming at each target frequency band;
detecting whether the wireless signal to be detected belongs to a target type according to the extracted frequency spectrum characteristics;
the target frequency band and the characteristic types of the target frequency band are determined in the following modes:
obtaining a first sample wireless signal, wherein the first sample wireless signal comprises: a wireless signal belonging to a target type and a wireless signal not belonging to the target type;
extracting the spectrum characteristic of the first sample wireless signal as a first sample spectrum characteristic;
obtaining a pearson correlation coefficient corresponding to each cluster of spectrum features in the first sample spectrum feature, wherein the spectrum features contained in each cluster of spectrum features are the same in frequency band and the same in type, and the pearson correlation coefficient corresponding to each cluster of spectrum features is used for characterizing: the frequency band of the frequency spectrum features is in the cluster of frequency spectrum features, and the type of the frequency spectrum features is the confidence coefficient that the wireless signals of the type to which the cluster of frequency spectrum features belongs belong to the target type;
preferentially selecting the Pearson correlation coefficient with high represented confidence from the obtained Pearson correlation coefficients;
aiming at each selected Pearson correlation coefficient, taking the frequency band of the spectrum feature in the spectrum feature cluster corresponding to the Pearson correlation coefficient as a target frequency band, and taking the type of the spectrum feature as the feature type of the target frequency band;
the Pearson correlation coefficient corresponding to each cluster of frequency spectrum features is calculated by adopting the following formula:
Figure FDA0002528374340000011
where ρ iskPearson's correlation coefficient, x, for the kth spectral feature clusterkIs the spectral feature of the kth spectral feature cluster, y is a flag indicating whether the wireless signal in the first sample wireless signal belongs to the target type in the form of a zero,
Figure FDA0002528374340000021
is xkThe average value of (a) of (b),
Figure FDA0002528374340000022
is the average value of the values of y,
Figure FDA0002528374340000023
is xkStandard deviation of (a)yIs the standard deviation of y, xi,kIs the k-th spectral characteristic, y, of the i-th wireless signal in the first sample wireless signaliN is the number of wireless signals in the first sample wireless signal, in order to indicate the mark whether the ith wireless signal in the first sample wireless signal belongs to the target type in a zero form.
2. The method according to claim 1, wherein said extracting the spectral feature of the first sample wireless signal as a first sample spectral feature comprises:
respectively extracting the power spectral density of each first sample wireless signal on a preset frequency point in a preset frequency band;
respectively calculating the spectrum characteristics of each first sample wireless signal on each sub-band of the preset frequency band based on the extracted power spectral density to serve as the spectrum characteristics of the sub-band;
aiming at each sub-band spectrum feature, obtaining a two-dimensional vector formed by the sub-band spectrum feature and a zero mark corresponding to the sub-band spectrum feature, wherein the zero mark corresponding to each sub-band spectrum feature is as follows: a flag, expressed in the form of a zero, indicating whether the first sample radio signal whose spectral characteristics include those of the sub-band belongs to the target type;
and aggregating the obtained two-dimensional vectors to obtain the first sample spectrum characteristic.
3. The method of claim 1, wherein the preferentially selecting the pearson correlation coefficients with high characterized confidence from the obtained pearson correlation coefficients comprises:
traversing and selecting each obtained Pearson correlation coefficient according to the following mode, and determining the Pearson correlation coefficient with high represented confidence coefficient in the obtained Pearson correlation coefficients:
selecting the largest not selected Pearson correlation coefficient from the obtained Pearson correlation coefficients as the current Pearson correlation coefficient;
adding the spectrum feature cluster corresponding to the current Pearson correlation coefficient to a traversal feature set;
taking the labeling information corresponding to each spectrum feature in the traversal feature set and the traversal feature set as the input of a first quadratic logistic regression model, selecting a model parameter which enables the accuracy of the detection result of the first quadratic logistic regression model to be highest based on a ten-fold cross-validation method, and finishing the training of the first quadratic logistic regression model, wherein the labeling information corresponding to each spectrum feature represents: whether the first sample wireless signal with the spectrum characteristic is a target type wireless signal or not, wherein the detection result of the first two-term logistic regression model is used for representing that: the frequency spectrum characteristic is whether the wireless signal of each frequency spectrum characteristic in the traversal characteristic set belongs to a target type;
judging whether the accuracy of the detection result corresponding to the selected model parameter is greater than the accuracy of the historical detection result;
if not, deleting the spectrum feature cluster corresponding to the current Pearson correlation coefficient from the traversal feature set;
and if so, determining that the current Pearson correlation coefficient is one of the obtained Pearson correlation coefficients with high represented confidence coefficient.
4. The method according to any one of claims 1 to 3, wherein the characteristic category of the target frequency band comprises at least one of the following characteristic categories:
mean of power spectral density;
variance of power spectral density;
differential characterization of power spectral density.
5. The method according to any one of claims 1-3, wherein the detecting whether the wireless signal to be detected belongs to a target type according to the extracted spectral features comprises:
inputting the extracted spectral features into a classifier obtained by pre-training to obtain a detection result of whether the wireless signal to be detected belongs to a target type;
wherein the two classifiers are obtained by training in the following way:
obtaining a second sampled wireless signal, wherein the second sampled wireless signal comprises: a wireless signal belonging to a target type and a wireless signal not belonging to the target type;
extracting the frequency spectrum characteristics of the second sample wireless signals on the target frequency band, belonging to the characteristic type of the target frequency band, as second sample frequency spectrum characteristics aiming at each target frequency band;
for each second sample wireless signal, obtaining a feature vector formed by the spectrum features belonging to the second sample wireless signal in the second sample spectrum features;
and constructing a second-term logistic regression model by taking the obtained feature vectors and a zero mark corresponding to each feature vector as known quantities and taking the weight parameter vectors and the bias parameters as unknown quantities, wherein the zero mark corresponding to each feature vector is as follows: a flag, expressed in the form of a zero, indicating whether the second sample wireless signal corresponding to each feature vector belongs to the target type;
obtaining the second logistic regression model to enable the maximum likelihood algorithm to obtain the value of the weight parameter and the value of the bias parameter when the maximum value is obtained;
and determining a second-term logistic regression model when the values of the weight parameter vector and the bias parameter are respectively obtained values as the two classifiers.
6. The method according to claim 5, wherein the constructing a second logistic regression model with the obtained eigenvectors and a zero flag corresponding to each eigenvector as known quantities and the weight parameter vectors and the bias parameters as unknown quantities comprises:
establishing the second term logistic regression model by adopting the following formula:
Figure FDA0002528374340000041
wherein x isiIs the i-th feature vector, yiA zero mark corresponding to the ith feature vector, ω is the weight parameter vector, and b is the bias parameter.
7. The method of claim 5, wherein obtaining the second logistic regression model such that the weight parameter value and the bias parameter value when the maximum likelihood algorithm obtains the maximum value comprises:
obtaining the second binomial logistic regression model by adopting a gradient descent method to calculate the maximum value of a preset log-likelihood function, so that the weight parameter value and the bias parameter value are obtained when the maximum value is obtained by a maximum likelihood algorithm;
wherein the preset log-likelihood function is:
Figure FDA0002528374340000042
wherein L (ω) is a log-likelihood function, xiIs the i-th feature vector, yiA zero mark corresponding to the ith eigenvector, N is the number of eigenvectors, omega is the weight parameter vector, and b is the bias parameter.
8. The method of claim 5, wherein determining a second-term logistic regression model as the second classifier when the values of the weight parameter vector and the bias parameter are obtained values respectively comprises:
determining a classifier represented by the following expression as the two classifiers:
Figure FDA0002528374340000051
wherein,
Figure FDA0002528374340000052
the value of the weight parameter vector is taken as the value,
Figure FDA0002528374340000053
taking the value of the bias parameter, wherein x is a vector which is input into the two classifiers and is formed by spectrum features, y is a zero mark output by the two classifiers, and the zero mark output by the two classifiers is: and a flag, expressed in the form of a zero, indicating whether the wireless signal represented by the vector input to the two classifiers belongs to the target type.
9. A wireless signal detection device, comprising:
the signal acquisition module is used for acquiring a wireless signal to be detected;
the frequency band obtaining module is used for obtaining a target frequency band of the spectral features to be extracted and obtaining the type of the spectral features to be extracted on each target frequency band as the characteristic type of each target frequency band;
the characteristic extraction module is used for extracting the frequency spectrum characteristics of the wireless signal to be detected, which belong to the characteristic category of the target frequency band on the target frequency band, aiming at each target frequency band;
the signal detection module is used for detecting whether the wireless signal to be detected belongs to a target type according to the extracted frequency spectrum characteristics;
wherein, the frequency band obtaining module includes:
a sample signal acquisition submodule for obtaining a first sample wireless signal, wherein the first sample wireless signal comprises: a wireless signal belonging to a target type and a wireless signal not belonging to the target type;
the sample characteristic extraction submodule is used for extracting the frequency spectrum characteristic of the first sample wireless signal as a first sample frequency spectrum characteristic;
a pearson correlation coefficient obtaining sub-module, configured to obtain a pearson correlation coefficient corresponding to each cluster of spectral features in the first sample spectral feature, where the spectral features included in each cluster are the same in frequency band and the types of the spectral features are the same, and the pearson correlation coefficient corresponding to each cluster of spectral features is used for characterizing: the frequency band of the frequency spectrum features is in the cluster of frequency spectrum features, and the type of the frequency spectrum features is the confidence coefficient that the wireless signals of the type to which the cluster of frequency spectrum features belongs belong to the target type; the Pearson correlation coefficient corresponding to each cluster of frequency spectrum features is calculated by adopting the following formula:
Figure FDA0002528374340000061
where ρ iskPearson's correlation coefficient, x, for the kth spectral feature clusterkIs the spectral feature of the kth spectral feature cluster, y is a flag indicating whether the wireless signal in the first sample wireless signal belongs to the target type in the form of a zero,
Figure FDA0002528374340000062
is xkThe average value of (a) of (b),
Figure FDA0002528374340000063
is the average value of the values of y,
Figure FDA0002528374340000064
is xkStandard deviation of (a)yIs the standard deviation of y, xi,kIs the k-th spectral characteristic, y, of the i-th wireless signal in the first sample wireless signaliA mark which represents whether the ith wireless signal in the first sample wireless signal belongs to the target type in a zero form, wherein n is the number of the wireless signals in the first sample wireless signal;
the pearson correlation coefficient optimization submodule is used for preferentially selecting the pearson correlation coefficient with high represented confidence from the obtained pearson correlation coefficients;
and the target frequency band determining submodule is used for taking the frequency band of the spectrum features in the spectrum feature cluster corresponding to the Pearson correlation coefficient as a target frequency band and taking the types of the spectrum features as the feature types of the target frequency band aiming at each selected Pearson correlation coefficient.
10. An electronic device is characterized by comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus;
a memory for storing a computer program;
a processor for implementing the method steps of any of claims 1 to 8 when executing a program stored in the memory.
CN201810723748.2A 2018-07-04 2018-07-04 Wireless signal detection method and device Active CN109104257B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810723748.2A CN109104257B (en) 2018-07-04 2018-07-04 Wireless signal detection method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810723748.2A CN109104257B (en) 2018-07-04 2018-07-04 Wireless signal detection method and device

Publications (2)

Publication Number Publication Date
CN109104257A CN109104257A (en) 2018-12-28
CN109104257B true CN109104257B (en) 2020-08-21

Family

ID=64845643

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810723748.2A Active CN109104257B (en) 2018-07-04 2018-07-04 Wireless signal detection method and device

Country Status (1)

Country Link
CN (1) CN109104257B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110518924B (en) * 2019-09-12 2021-04-09 北京维普无限智能技术有限公司 Method for improving compatibility of wireless signal transceiver
EP3825703B1 (en) * 2019-11-20 2023-08-30 Rohde & Schwarz GmbH & Co. KG Method and system for detecting and/or classifying a wanted signal
CN113300986B (en) * 2021-04-17 2022-06-24 湖南红船科技有限公司 Unmanned aerial vehicle image transmission signal and hotspot signal identification method, medium and computer equipment
CN114978410B (en) * 2022-07-28 2022-11-18 杭州恒生数字设备科技有限公司 Method and system for shielding radio signal

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102364885B (en) * 2011-10-11 2014-02-05 宁波大学 Frequency spectrum sensing method based on signal frequency spectrum envelope
US9094080B1 (en) * 2013-01-04 2015-07-28 Marvell International Ltd. Method and apparatus for estimating statistics in wireless systems
CN104468001B (en) * 2014-11-26 2017-04-19 北京邮电大学 Signal identification method and system based on radio signal frequency spectrum feature template
CN107222875B (en) * 2017-07-10 2019-10-01 东南大学 Small base station switch control method based on big data analysis

Also Published As

Publication number Publication date
CN109104257A (en) 2018-12-28

Similar Documents

Publication Publication Date Title
CN109104257B (en) Wireless signal detection method and device
CN110222170B (en) Method, device, storage medium and computer equipment for identifying sensitive data
US11057788B2 (en) Method and system for abnormal value detection in LTE network
Lapuschkin et al. Analyzing classifiers: Fisher vectors and deep neural networks
CN110197205B (en) Image identification method of multi-feature-source residual error network
US20190166024A1 (en) Network anomaly analysis apparatus, method, and non-transitory computer readable storage medium thereof
CN108447057B (en) SAR image change detection method based on significance and depth convolution network
WO2019051941A1 (en) Method, apparatus and device for identifying vehicle type, and computer-readable storage medium
JP2017531410A5 (en)
CN111967535B (en) Fault diagnosis method and device for temperature sensor of grain storage management scene
CN113269706B (en) Laser radar image quality evaluation method, device, equipment and storage medium
CN109002810A (en) Model evaluation method, Radar Signal Recognition method and corresponding intrument
CN106452934B (en) Method and device for analyzing network performance index change trend
CN109525994A (en) High energy efficiency frequency spectrum sensing method based on support vector machines
TWI775186B (en) Rf fingerprint signal processing device and rf fingerprint signal processing method
CN104820841A (en) Hyper-spectral classification method based on low-order mutual information and spectral context band selection
CN108470194B (en) Feature screening method and device
CN107247873A (en) A kind of recognition methods of differential methylation site
CN105184286A (en) Vehicle detection method and detection device
CN112560971A (en) Image classification method and system for active learning self-iteration
CN113052577A (en) Method and system for estimating category of virtual address of block chain digital currency
CN117727308B (en) Mixed bird song recognition method based on deep migration learning
CN113837255B (en) Method, apparatus and medium for predicting cell-based antibody karyotype class
Bach et al. Analyzing classifiers: Fisher vectors and deep neural networks
CN105743756A (en) Frame detection method based on Adaboost algorithm in Wi-Fi 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