CN104868962A - Spectrum detection method and device based on compressed sensing - Google Patents

Abstract

The invention discloses a spectrum detection method and device based on compressed sensing. The method comprises the following steps: performing compressed sampling on signals in a spectrum environment based on a sampling rate to obtain compressed sampling signals; performing recovery with the compressed sampling signals to obtain a cyclic spectrum; and detecting whether or not a spectrum hole exists in the spectrum environment according to the cyclic spectrum. Through adoption of the method, the required sampling rate during spectrum detection is lowered effectively, and high-accuracy spectrum hole detection is realized under the condition of low sampling rate. Meanwhile, the reconstruction process of a Nyquist sampling signal is omitted, and real-time identification is performed on a symbol rate and a modulation way, so that the complexity of a scheme and the sampling rate required by spectrum detection are lowered, and the compression gain is increased.

Description

Based on frequency spectrum detecting method and the device of compressed sensing

Technical field

The present invention relates to wireless communication field, particularly relate to cognition wireless electrical domain, specifically, relate to a kind of frequency spectrum detecting method based on compressed sensing and device.

Background technology

Along with the development of wireless communication technology, the demand of radio communication service to bandwidth is more and more higher.Therefore, higher message transmission rate and wider frequency spectrum resource is needed badly.Nowadays fixing wireless frequency spectrum distribution policy makes the utilance of some frequency range not high and causes the free time, only has some to use higher than frequency range utilances such as mobile communication more frequently.Cognitive radio technology can make secondary user's dynamically utilize not by the frequency spectrum resource that primary user takies, and alleviation frequency spectrum resource is nervous and the availability of frequency spectrum is low and the contradiction of depositing.

Frequency spectrum detection can obtain the busy-idle condition of primary user, is the basis of cognitive radio, therefore receives and pay close attention to widely.Frequency spectrum detecting method is divided into non-cooperative detection, cooperative detection and the detection based on interference temperature.Wherein, non-cooperative detection detects also referred to as transmitter, is a kind of frequency spectrum detecting method be most widely used, and this detection method carries out frequency spectrum detection by whether there is master transmitter signal in measurement environment.Non-cooperative detection comprises matched filter detection, energy measuring and cyclo-stationary further and detects.Wherein, cyclo-stationary detects and still have higher detection perform under the condition not needing the priori knowing primary user.

Mainly there is following problem in existing frequency spectrum detection technology, spectral analysis algorithm is complicated, needs higher sampling rate.One section of patent No. is that the patent of CN103873170A proposes compressed sensing to be used for frequency spectrum detection.Compressed sensing is also referred to as compression sampling, and based on the openness proposition of signal, it can be sampled to signal under far below the condition of Nyquist sampling rate do not reconstruct nyquist sampling signal loss of information.But due to needs reconstruct nyquist sampling signal, so compressed sensing algorithm has higher complexity usually.

To sum up, need a kind of requirement that both can meet to real-time in cognition wireless electrical domain badly, the frequency spectrum detecting method can effectively implemented again is to solve the problem.

Summary of the invention

One of technical problem to be solved by this invention needs to provide a kind of requirement that both can meet to real-time in cognition wireless electrical domain, the frequency spectrum detecting method can effectively implemented again.

In order to solve the problems of the technologies described above, the embodiment of the application provide firstly a kind of frequency spectrum detecting method based on compressed sensing, comprises, and carries out compression sampling obtain compression sampling signal based on sampling rate to the signal in spectrum environment; Described compression sampling signal recuperation is utilized to obtain Cyclic Spectrum; Detect in spectrum environment whether there is frequency spectrum cavity-pocket according to described Cyclic Spectrum.

Preferably, the method also comprises the character rate and the modulation system that identify compression sampling signal based on described Cyclic Spectrum, and adjusts sampling rate in real time according to the detection probability of described character rate, described modulation system and setting.

Preferably, identify the character rate of compression sampling signal and the step of modulation system based on Cyclic Spectrum, comprising: the Cyclic Spectrum by compression sampling signal recuperation is projected to circulation frequency domain to obtain the circulation spectral envelope of compression sampling signal; Carrier frequency and the character rate of compression sampling signal is extracted according to described circulation spectral envelope; Utilize described carrier frequency and character rate to produce template signal, and described template signal is projected to circulation frequency domain to obtain the circulation spectral envelope of template signal; The modulation system of template signal minimum for vector angle in the inner product of vectors of the circulation spectral envelope of compression sampling signal and the circulation spectral envelope of template signal is judged to be the modulation system of compression sampling signal.

Preferably, Cyclic Spectrum is recovered according to following formula:

$s_x^{\left(c\right)}=H^{+}{r}_y$

In formula, r _yfor all row in the covariance matrix of compression sampling signal being coupled together the vector formed, for all row in Cyclic Spectrum matrix being coupled together the vector formed; H ⁺for the MoorePenrose generalized inverse matrix of transformation matrix H; Wherein, H=Λ T ⁺, T ⁺for the Moore Penrose generalized inverse matrix of T, matrix $\mathrm{\Λ}=Q_M(\mathrm{\Φ}\⊗\mathrm{\Φ})/B_p,$ Φ is condensation matrix, matrix $T=(I\⊗F)\underset{v=0}{\overset{N-1}{\mathrm{\Σ}}}(D_v^T\⊗G_v)/B_p,$ I is that N × N ties up unit matrix, and F is fast Fourier transform matrix, Q _m, B _p, G _vand D _vbe transformation matrix, for Kronecker amasss.

Preferably, in real time the step of adjustment sampling rate specifically comprises: the compression gains of detection probability determination compression sampling utilizing described character rate, described modulation system and setting; The compression ratio of compression sampling is obtained according to described compression gains, and by the sampling rate of described compression ratio and Nyquist sampling rate determination compression sampling.

On the other hand, additionally provide a kind of frequency spectrum detection device based on compressed sensing, comprising: compression sampling unit, it carries out compression sampling based on selected sampling rate to the signal in spectrum environment and obtains compression sampling signal; Spectral analysis unit, it utilizes described compression sampling signal recuperation Cyclic Spectrum; Cavity detecting unit, whether it exists frequency spectrum cavity-pocket according in described Cyclic Spectrum estimated spectral environment.

Preferably, frequency spectrum detection device based on compressed sensing also comprises pattern recognition unit and parameter adjustment unit, wherein, described pattern recognition unit identifies character rate and the modulation system of compression sampling signal based on described Cyclic Spectrum, and described parameter adjustment unit adjusts sampling rate in real time according to the detection probability of described modulation system, described character rate and setting.

Preferably, pattern recognition unit is according to the character rate of following steps identification compression sampling signal and modulation system: the Cyclic Spectrum by compression sampling signal recuperation is projected to circulation frequency domain to obtain the circulation spectral envelope of compression sampling signal; Carrier frequency and the character rate of compression sampling signal is extracted according to described circulation spectral envelope; Utilize described carrier frequency and character rate to produce template signal, and described template signal is projected to circulation frequency domain to obtain the circulation spectral envelope of template signal; The modulation system of template signal minimum for vector angle in the inner product of vectors of the circulation spectral envelope of compression sampling signal and the circulation spectral envelope of template signal is judged to be the modulation system of compression sampling signal.

Preferably, spectral analysis unit recovers Cyclic Spectrum according to following formula:

$s_x^{\left(c\right)}=H^{+}{r}_y$

In formula, r _yfor all row in the covariance matrix of compression sampling signal being coupled together the vector formed, for all row in Cyclic Spectrum matrix being coupled together the vector formed; H ⁺for the MoorePenrose generalized inverse matrix of transformation matrix H; Wherein, H=Λ T ⁺, T ⁺for the Moore Penrose generalized inverse matrix of T, matrix $\mathrm{\Λ}=Q_M(\mathrm{\Φ}\⊗\mathrm{\Φ})/B_p,$ Φ is condensation matrix, matrix $T=(I\⊗F)\underset{v=0}{\overset{N-1}{\mathrm{\Σ}}}(D_v^T\⊗G_v)/B_p,$ I is that N × N ties up unit matrix, and F is fast Fourier transform matrix, Q _m, B _p, G _vand D _vbe transformation matrix, for Kronecker amasss.

Preferably, parameter adjustment unit adjusts sampling rate in real time according to following steps: the compression gains of detection probability determination compression sampling utilizing described character rate, described modulation system and setting; The compression ratio of compression sampling is obtained according to described compression gains, and by the sampling rate of described compression ratio and Nyquist sampling rate determination compression sampling.

Compared with prior art, the one or more embodiments in such scheme can have the following advantages or beneficial effect by tool:

Detecting by compressed sensing being applied to cyclo-stationary, sampling rate required when significantly reducing frequency spectrum detection, achieving the frequency spectrum cavity-pocket of high accuracy under low sampling rate condition and detecting.Simultaneously by eliminate nyquist sampling signal restructuring procedure and by carrying out Real time identification to character rate and modulation system, reduce the complexity of the complexity of scheme and the sample rate needed for frequency spectrum detection and scheme, improve compression gains.

Other advantages of the present invention, target, to set forth in the following description to a certain extent with feature, and to a certain extent, based on will be apparent to those skilled in the art to investigating hereafter, or can be instructed from the practice of the present invention.Target of the present invention and other advantages can by specifications below, claims, and in accompanying drawing, specifically noted structure realizes and obtains.

Accompanying drawing explanation

Accompanying drawing is used to provide the further understanding of technical scheme to the application or prior art, and forms a part for specification.Wherein, the expression accompanying drawing of the embodiment of the present application and the embodiment one of the application are used from the technical scheme explaining the application, but do not form the restriction to technical scheme.

Fig. 1 is the schematic flow sheet of the frequency spectrum detecting method of the embodiment of the present application;

The relation schematic diagram that Fig. 2 (a) is compression gains and modulation system, the relation schematic diagram that Fig. 2 (b) is compression gains and character rate;

Fig. 3 is the schematic flow sheet of the frequency spectrum detecting method of the embodiment of the present application;

Fig. 4 is the character rate identification of the embodiment of the present application and the schematic flow sheet of Modulation Mode Recognition method;

Fig. 5 is the performance simulation schematic diagram of the Modulation Mode Recognition method of the embodiment of the present application;

Fig. 6 is the relation schematic diagram of detection probability and compression gains under different modulation system;

Fig. 7 is the relation schematic diagram of detection probability and compression gains under different character rate;

Fig. 8 is the structural representation of the frequency spectrum detection device of the embodiment of the present application.

Embodiment

Describe embodiments of the present invention in detail below with reference to drawings and Examples, to the present invention, how application technology means solve technical problem whereby, and the implementation procedure reaching relevant art effect can fully understand and implement according to this.Each feature in the embodiment of the present application and embodiment, can be combined with each other under prerequisite of not conflicting mutually, the technical scheme formed is all within protection scope of the present invention.

Compressed sensing detects with cyclo-stationary and combines for frequency spectrum detection by the embodiment of the application.The higher frequency spectrum detection performance utilizing cyclo-stationary to detect on the one hand, on the other hand, compressed sensing can reduce the requirement of the sampling rate to front end analogue sample devices effectively, simultaneously owing to decreasing the data volume that sampling obtains, reduce the demand to transmission and memory device, so saved cost.Therefore, validity and the exploitativeness of frequency spectrum detection is improve based on the cyclo-stationary detection method of compressed sensing.

Cyclo-stationary detects the characteristic being carried out analytical cycle stationary signal by Spectral correlation function (spectral coherence function, SCF), has higher complexity.Compressed sensing needs to carry out frequency spectrum detection according to compression sampling signal reconstruction nyquist sampling signal recycling cyclo-stationary detection algorithm, also has higher complexity.Therefore the cyclo-stationary detection based on compressed sensing has very high complexity.High complexity can affect the real-time of frequency spectrum detection, thus affects it in the application to the higher cognition wireless electrical domain of requirement of real-time.In order to improve the real-time of frequency spectrum detection further, in the embodiment of the application, the cyclo-stationary detection algorithm based on compressed sensing being simplified, obtaining good real-time.Be described in detail below with reference to accompanying drawing.

Fig. 1 is the schematic flow sheet of the frequency spectrum detecting method of the embodiment of the present application.As shown in Figure 1, the frequency spectrum detecting method of the embodiment of the present application comprises the following steps: step S110, carry out compression sampling obtain compression sampling signal based on sampling rate to the signal in spectrum environment; Step S120, described compression sampling signal recuperation is utilized to obtain Cyclic Spectrum; Step S130, detect in spectrum environment whether there is frequency spectrum cavity-pocket according to described Cyclic Spectrum.

Carry out compression sampling to the signal in spectrum environment to carry out according to expression formula (1):

Y＝ΦX (1)

In formula, Φ is the condensation matrix that M × N (M<<N) ties up, and Y represents compression sampling signal, and X represents nyquist sampling signal, and its sparse expression form is as shown in expression formula (2):

X＝ΨS (2)

In formula, Ψ is sparse observing matrix, and S is the sparse expression of X.Can find out, by compression sampling, the data volume that sampling can be made to obtain reduces greatly.

Cyclo-stationary of the prior art detects and carries out based on nyquist sampling signal, thus compression sampling and cyclo-stationary are detected combine time, need the reconstruct carrying out nyquist sampling signal.Being specially, first recovering S, by solving convex optimization problem as shown in expression formula (3) by compression sampling signal Y:

$\begin{array}{c}\underset{s}{\min }{\left|\right|S\left|\right|}_1\\ s.t.\mathrm{\&Phi;\&Psi;S}=Y\end{array}---\left(3\right)$

In formula, || S|| ₁represent the l of S ₁norm.Utilization minimizes l ₁norm recovers S, then the S recovering to obtain is substituted into the reconstruct that expression formula (2) just can realize nyquist sampling signal X.If X time become covariance r _x(n, v)=E{x (nT _s) x (nT _s+ vT _s) meet equation r _x(n, v)=r _x(n+kT ₀, v) (wherein T ₀≠ 0), then X is cyclo-stationary.

X is carried out Fourier expansion, and the circulation covariance of X is as shown in expression formula (4):

${{\tilde{r}}_x}^{\left(c\right)}(a,v)=\left\{\frac{1}{N}\underset{n=0}{\overset{N-1-v}{\mathrm{\&Sigma;}}}{r}_x(n,v)e^{-j\frac{2\mathrm{\&pi;}}{N}\mathrm{an}}\right\}{e}^{-j\frac{\mathrm{\&pi;}}{N}\mathrm{av}}---\left(4\right)$

In formula, parameter when N is Fourier expansion, represents the number of cycle frequency herein; for the circulation covariance of X, wherein, a is the parameter relevant to cycle frequency, and v is the discretization result in time interval when carrying out correlation computations; r _xbecome covariance when (n, v) is X, wherein, n is the sampling instant after discretization.

Fourier expansion is carried out to circulation covariance and can obtain Cyclic Spectrum SCF, as shown in expression formula (5):

$s_x^{\left(c\right)}(a,b)=\underset{v=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{{\tilde{r}}_x}^{\left(c\right)}(a,v)e^{-j\frac{2\mathrm{\&pi;}}{N}(b-\frac{N-1}{2})v}---\left(5\right)$

In formula, for Cyclic Spectrum, wherein, a is the parameter relevant to cycle frequency, and b is the parameter with frequency dependence.

Cyclo-stationary detects and utilizes the Cyclic Spectrum obtained to estimate frequency spectrum cavity-pocket.The restructuring procedure being carried out nyquist sampling signal by expression formula (1), (2), (3) has higher complexity, so save above-mentioned restructuring procedure in the embodiment of the application, directly utilize compression sampling signal recuperation Cyclic Spectrum, thus significantly reduce the complexity of algorithm.Detailed process is as follows:

By time become covariance be R with matrix representation, as shown in expression formula (6):

When nyquist sampling signal is real signal, R is symmetrical matrix, can be expressed as the vector form shown in expression formula (7):

r _x＝[r _x(0,0),r _x(1,0),...,r _x(N-1,0),r _x(0,1),r _x(1,1),...,r _x(N-2,1),...,r _x(0,N-1)] ^T(7)

In formula, r _xfor all row of R being coupled together the vector formed, by r _xr is designated as with the relation of R _x=B _pvec{R}, wherein, B _pfor transformation matrix, vec{} represents the vector being coupled together by all row of matrix and formed.To be circulated covariance and Cyclic Spectrum same matrix representation is and and both can by Fourier transform and time the covariance matrix R that becomes connect, as shown in expression formula (8):

$\begin{array}{c}{\tilde{R}}_x^{\left(c\right)}=\underset{v=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{G}_v{\mathrm{RD}}_v\\ S_x^{\left(c\right)}={\tilde{R}}_x^{\left(c\right)}F\end{array}---\left(8\right)$

In formula, G _vand D _vbe transformation matrix, d _vfor N × N ties up matrix, the individual element of this matrix (v, v) is 1, and it is location parameter that all the other elements are 0, v.

represent N rank fast Fourier transform matrix.

Order expression formula (9) can be obtained:

$s_x^{\left(c\right)}=\mathrm{vec}\left\{{\tilde{R}}_x^{\left(c\right)}F\right\}=(I\&CircleTimes;F)\underset{v=0}{\overset{N-1}{\mathrm{\&Sigma;}}}(D_v^T\&CircleTimes;G_v)\mathrm{vec}\left\{R\right\}={\mathrm{Tr}}_x---\left(9\right)$

In formula, I is the unit matrix of N × N dimension, represent that Kronecker amasss.The relation that what expression formula (9) represented is between Cyclic Spectrum and nyquist sampling signal.

Further, compression sampling signal time become covariance matrix into according to method above, by R _ybe expressed as vector form, as shown in expression formula (10):

r _y＝Q _Mvec{R _y} (10)

In formula, $Q_M\&Element;{\{0,\frac{1}{2},1\}}^{\frac{M(M+1)}{2}\&times;M^2}$ For transformation matrix.

The covariance matrix R of compression sampling signal can be obtained according to expression formula (1) _ywith the covariance matrix R of nyquist sampling signal _xrelation as shown in expression formula (11):

R _y＝ΦR _xΦ ^T(11)

R can be obtained by expression formula (10) and expression formula (11) _ywith r _xrelation as shown in expression formula (12):

$r_y=Q_M\mathrm{vec}\left\{{\mathrm{\&Phi;R}}_x{\mathrm{\&Phi;}}^T\right\}=Q_M(\mathrm{\&Phi;}\&CircleTimes;\mathrm{\&Phi;})\mathrm{vec}\left\{R\right\}={\mathrm{\&Lambda;r}}_x---\left(12\right)$

In formula, $\mathrm{\&Lambda;}=Q_M(\mathrm{\&Phi;}\&CircleTimes;\mathrm{\&Phi;})/B_p$

The linear relationship of compression sampling matrix and Cyclic Spectrum matrix can be obtained as shown in expression formula (13) by expression formula (9) and expression formula (12):

$r_y={\mathrm{\&Lambda;T}}^{+}{s}_x^{\left(c\right)}={\mathrm{Hs}}_x^{\left(c\right)}---\left(13\right)$

In formula, T ⁺for the Moore Penrose generalized inverse matrix of T.The both sides of above-mentioned expression formula are multiplied by the generalized inverse matrix of matrix H simultaneously, expression formula (14) can be obtained:

$s_x^{\left(c\right)}=H^{+}{r}_y---\left(14\right)$

Directly Cyclic Spectrum can be obtained by expression formula (14) by compression sampling signal in the implementation of algorithm, and without the need to carrying out the reconstruct of nyquist sampling signal, thus under the condition not affecting frequency spectrum detection performance, reduce the complexity of algorithm.

After obtaining Cyclic Spectrum, entropy estimate algorithm can be utilized to estimate frequency spectrum cavity-pocket.Be specially, first Cyclic Spectrum SCF be mapped to circulation frequency domain, as shown in expression formula (15):

$S\left(\mathrm{\&alpha;}\right)=\underset{f_i}{\max }\left(S_x^{\left(c\right)}(f,\mathrm{\&alpha;})\right)---\left(15\right)$

In formula, $\mathrm{\&alpha;}=\frac{1}{{\mathrm{NT}}_s}a$ For cycle frequency, $f=\frac{1}{{\mathrm{NT}}_s}(b-\frac{N-1}{2})\&Element;(-\frac{f_s}{2},\frac{f_s}{2})$ For frequency.Recycling expression formula (16) calculates peak factor Γ _r:

${\mathrm{\&Gamma;}}_R=\max \left(S\left(\mathrm{\&alpha;}\right)\right)/\sqrt{\left(\underset{\mathrm{\&alpha;}=0}{\overset{N}{\mathrm{\&Sigma;}}}{S}^2\left(\mathrm{\&alpha;}\right)\right)/N}---\left(16\right)$

In formula, N represents the number of the cycle frequency chosen.Then the presence or absence of frequency spectrum cavity-pocket is judged by expression formula (17):

$\left\{\begin{array}{c}{H}_0:{\mathrm{\&Gamma;}}_R\&le;{\mathrm{\&Gamma;}}_{\mathrm{TH}}\\ H_1:{\mathrm{\&Gamma;}}_R>{\mathrm{\&Gamma;}}_{\mathrm{TH}}\end{array}\right.---\left(17\right)$

In formula, Γ _tHfor the thresholding preset.When being judged as H ₀time, represent that frequency spectrum cavity-pocket exists, when being judged as H ₁time, represent that primary user exists.

Directly obtain Cyclic Spectrum by compression sampling signal, and without the need to carrying out the reconstruct of nyquist sampling signal, under the condition not affecting frequency spectrum detection performance, reduce the complexity of algorithm.

In other embodiments of the application, adjusting sampling rate by carrying out identification to modulation system and character rate, reduce further the compression ratio needed for compressed sensing, improve compression gains.

Compression ratio needed for compressed sensing is determined by signal degree of rarefication and detection bandwidth, and the factor therefore affecting signal degree of rarefication will affect the compression ratio needed for compressed sensing.The degree of rarefication of signal comprises inner degree of rarefication and outside degree of rarefication.Inside degree of rarefication and the signal madulation mode of signal are closely related, and outside degree of rarefication and character rate are closely related.So the compression ratio needed for compressed sensing is subject to the impact of modulation system and character rate, the compression gains that namely compressed sensing obtains can be subject to the impact of modulation system and character rate.

The relation schematic diagram that Fig. 2 (a) is compression gains and modulation system, the relation schematic diagram that Fig. 2 (b) is compression gains and character rate.Abscissa in Fig. 2 is compression gains G, can be determined by expression formula (18):

G＝1-C _r＝1-M/N (18)

In formula, C _rrepresent compression ratio, M represents compression sampling speed, and N represents Nyquist sampling rate.

Ordinate in Fig. 2 is normalized mean squared error MSE, can be defined by expression formula (19):

In formula, represent the Cyclic Spectrum utilizing compression sampling signal recuperation to obtain, SCF represents the Cyclic Spectrum utilizing nyquist sampling signal to obtain.

Normalized mean square error MSE can show the frequency spectrum detection loss of energy that compressed sensing is brought, namely can be used for represent Cyclic Spectrum recover performance.As can be seen from Figure 2, normalized mean squared error MSE can increase with the increase of compression gains G, and the performance namely detected when the compression gains time-frequency spectrum increasing compressed sensing can decline.Further, in Fig. 2 (a), when identical MSE, the compression gains that different modulation systems obtains is different; In Fig. 2 (b), when identical MSE, the compression gains that different character rates obtains also is different, and the compression gains of the higher acquisition of character rate is less.Therefore, in the embodiment of the application, by determining the compression ratio of compressed sensing to the identification of modulation system and character rate, effectively can reduce sampling rate, and improve compression gains.

Fig. 3 is the schematic flow sheet of the frequency spectrum detecting method of the embodiment of the present application.This detection method comprises: step S310, carry out compression sampling obtain compression sampling signal based on by adjusting the sampling rate obtained in real time the signal in spectrum environment; Step S320, utilize described compression sampling signal recuperation Cyclic Spectrum, and whether there is frequency spectrum cavity-pocket according in Cyclic Spectrum estimated spectral environment; Step S330, the character rate identification carrying out compression sampling signal based on described Cyclic Spectrum and Modulation Mode Recognition; Step S340, adjust sampling rate in real time according to the detection probability of described modulation system, described character rate and setting.Wherein, step S320 and step S330 and step S340 can carry out simultaneously.Step S310 is identical with step S110, and step S320 is identical with step S130 with step S120, does not repeat them here.Step S330 and step S340 is described in detail below in conjunction with Fig. 4 and Fig. 5.

Fig. 4 is the character rate identification of the embodiment of the present application and the schematic flow sheet of Modulation Mode Recognition method.This recognition methods comprises the following steps: step S410, the Cyclic Spectrum by compression sampling signal recuperation is projected to circulation frequency domain to obtain the circulation spectral envelope of compression sampling signal; Step S420, extract carrier frequency and the character rate of compression sampling signal according to described circulation spectral envelope; Step S430, utilize described carrier frequency and character rate to produce template signal, and described template signal is projected to circulation frequency domain to obtain the circulation spectral envelope of template signal; Step S440, the modulation system of template signal minimum for vector angle in the inner product of vectors of the circulation spectral envelope of compression sampling signal and the circulation spectral envelope of template signal is judged to be the modulation system of compression sampling signal.

Particularly, the Cyclic Spectrum SCF by compression sampling signal recuperation is projected to circulation frequency domain to be realized by expression formula (15).

The process of Modulation Mode Recognition can be represented by the form shown in expression formula (20):

$M_n=\arg \underset{i}{\min }|<S\left(\mathrm{\&alpha;}\right),S_i\left(\mathrm{\&alpha;}\right)>|---\left(20\right)$

It should be noted that, in other embodiments of the application, the process that above-mentioned basis is identified character rate and modulation system by the Cyclic Spectrum of compression sampling signal recuperation, the method of Corpus--based Method feature or maximum likelihood method etc. can also be used to complete, the concrete recognition methods used is not limited.

Character rate identification and the Modulation Mode Recognition method of the embodiment of the present application have good recognition performance, and effectively can identify character rate and the modulation system of measured signal in low signal-to-noise ratio situation, reliability is high.Fig. 5 is the performance simulation schematic diagram of the Modulation Mode Recognition method of the embodiment of the present application.As can be seen from the figure, when signal to noise ratio reaches-8dB, Modulation Mode Recognition accuracy rate can reach 100%.

After the character rate obtaining signal in spectrum environment by identification and modulation system, can adjust in conjunction with the sampling rate of detection probability to compression sampling.Be specially, first the detection probability of setting is utilized and by identifying the modulation system of signal to be detected that obtains and the compression gains of character rate determination compression sampling, the compression ratio of compression sampling is obtained again according to compression gains, and according to the sampling rate of compression ratio and Nyquist sampling rate determination compression sampling.

Known by carrying out test to the data sample of training set, detection probability can reduce with the increase of compression gains, and under the condition of same detection probability, and compression gains will change along with the change of modulation system and character rate.Above-mentioned relation can see Fig. 6 and Fig. 7.

Fig. 6 is the relation schematic diagram of detection probability and compression gains under different modulation system, and Fig. 7 is the relation schematic diagram of detection probability and compression gains under different character rate, and two width figure are all measured by emulation experiment when false alarm probability is 0.05.In figure 6, ordinate represents detection probability Pd, and abscissa represents compression gains G, can find out, for different modulation systems, detection probability all reduces along with the increase of compression gains, and this also tallies with the actual situation.Further, for certain detection probability, its compression gains of different modulation systems is different, and modulation system is that the compression gains of 2PSK is minimum, and the compression gains of 4PSK is placed in the middle, and the compression gains of MSK is maximum.Select fixing detection probability due to general in practical application, the modulation system that therefore can obtain according to identification and character rate select suitable compression gains.For example, if the character rate in Fig. 6 is 5Hz, and be also 5Hz by the character rate identifying the detected signal obtained, and the modulation system identifying the detected signal obtained is MSK, so can determine that compression gains is 0.3 according to Fig. 6.And in prior art, when not identifying character rate and modulation system, generally sample with the modulation system of 2PSK, to ensure that the signal of no matter which kind of modulation system can both be sampled efficiently.As can be seen from the figure, when sampling with 2PSK, its compression gains is 0.19, and average compression gains can be improved 30% ([(0.3-0.19)/3+ (0.25-0.19)/3]/0.19*100%=30%) by the identification therefore by introducing modulation system.

Same, if the modulation system in Fig. 7 is MSK, and the modulation system identifying the detected signal obtained also is MSK, and is 2Hz by the character rate identifying the detected signal obtained, and so can determine that compression gains is 0.34 according to Fig. 7.And in prior art, when not identifying modulation system, generally sample with a higher character rate, to ensure that the signal of any speed can both be sampled efficiently.As can be seen from the figure, when sampling with 10Hz, its compression gains is 0.19, therefore average compression gains can be improved 35% ([(0.34-0.19)/3+ (0.24-0.19)/3]/0.19*100%=35%) by the identification of created symbol speed.

In use, the relation data of detection probability under the compound mode of multiple character rate and modulation system and compression gains can be obtained in advance according to training set data, for example, choose the modulation system often used, about 20 to 30 kinds.Character rate can be chosen at certain intervals, and about 20% of the optional current sign speed in interval, carries out permutation and combination by above-mentioned modulation system and character rate and obtain corresponding relation data.The following mode by data query is according to the compression gains of the relation data determination measured signal prestored, after the compression gains determining measured signal, can further according to expression formula (18), determine compression ratio, and compression ratio obtains according to the ratio of compression sampling speed and Nyquist sampling rate, therefore suitable compression sampling speed can be determined in conjunction with Nyquist sampling rate, if presently used sampling rate and this speed inconsistent, then adjust according to this speed.The method of the self-adaptative adjustment compression ratio based on modulation system and character rate identification that the embodiment of the present application adopts, compared to traditional compressed sensing algorithm, sample rate that can be lower obtains higher compression gains, and more excellent detection perform.

Fig. 8 is the structural representation of the frequency spectrum detection device of the embodiment of the present application.Can find out, this frequency spectrum detection device comprises, compression sampling unit 81, spectral analysis unit 82, empty detecting unit 83, pattern recognition unit 84 and parameter adjustment unit 85.

Compression sampling unit 81, is connected with parameter adjustment unit 85 with spectral analysis unit 82, for the input signal of the feedback signal and the external world that accept parameter adjustment unit 85, exports sampled result to spectral analysis unit 82 simultaneously.It carries out compression sampling based on selected sampling rate to the signal in spectrum environment and obtains compression sampling signal.

Spectral analysis unit 82, is connected with empty detecting unit 83 with compression sampling unit 81, and it utilizes described compression sampling signal recuperation Cyclic Spectrum, completes main computing function.On the other hand, be also connected with pattern recognition unit 84, for the signal recovering to obtain is real-time transmitted to pattern recognition unit 84.

Cavity detecting unit 83, whether it exists frequency spectrum cavity-pocket according in described Cyclic Spectrum estimated spectral environment.

Pattern recognition unit 84, is positioned on feedback path, is connected with spectral analysis unit 82, for obtaining input signal from it, is connected with parameter adjustment unit 85, for exporting recognition result to it.It mainly identifies character rate and the modulation system of compression sampling signal based on the Cyclic Spectrum obtained.

Parameter adjustment unit 85, be positioned on feedback path, be connected with pattern recognition unit 84, for receiving recognition result, be connected with compression sampling unit 81, for transmitting the sampling rate determined in real time, it mainly adjusts sampling rate in real time according to the detection probability of modulation system, character rate and setting.

The cyclo-stationary detection method based on compressed sensing of simplification that the present invention proposes, is applied to cyclo-stationary and detects by compressed sensing, achieve the frequency spectrum cavity-pocket of high accuracy under low sampling rate condition and detect.Meanwhile, the process eliminating nyquist sampling signal reconstruction by the derivation of equation reduces the complexity of scheme.In addition, propose a kind of compressed sensing algorithm based on modulation system and character rate identification, solve the problem that compression ratio is chosen, reduce further the complexity of sample rate needed for frequency spectrum detection and scheme, improve compression gains.

Although the execution mode disclosed by the present invention is as above, the execution mode that described content just adopts for the ease of understanding the present invention, and be not used to limit the present invention.Technical staff in any the technical field of the invention; under the prerequisite not departing from the spirit and scope disclosed by the present invention; any amendment and change can be done what implement in form and in details; but scope of patent protection of the present invention, the scope that still must define with appending claims is as the criterion.

Claims (10)

1., based on a frequency spectrum detecting method for compressed sensing, comprising:

Based on sampling rate, compression sampling is carried out to the signal in spectrum environment and obtain compression sampling signal;

Described compression sampling signal recuperation is utilized to obtain Cyclic Spectrum;

Detect in spectrum environment whether there is frequency spectrum cavity-pocket according to described Cyclic Spectrum.

2. method according to claim 1, is characterized in that, further comprising the steps of:

Identify character rate and the modulation system of compression sampling signal based on described Cyclic Spectrum, and adjust sampling rate in real time according to the detection probability of described character rate, described modulation system and setting.

3. method according to claim 2, is characterized in that, identifies the character rate of compression sampling signal and the step of modulation system, comprising based on described Cyclic Spectrum:

Cyclic Spectrum by compression sampling signal recuperation is projected to circulation frequency domain to obtain the circulation spectral envelope of compression sampling signal;

Carrier frequency and the character rate of compression sampling signal is extracted according to described circulation spectral envelope;

Utilize described carrier frequency and character rate to produce template signal, and described template signal is projected to circulation frequency domain to obtain the circulation spectral envelope of template signal;

The modulation system of template signal minimum for vector angle in the inner product of vectors of the circulation spectral envelope of compression sampling signal and the circulation spectral envelope of template signal is judged to be the modulation system of compression sampling signal.

4. according to the method in any one of claims 1 to 3, it is characterized in that, recover Cyclic Spectrum according to following formula:

$s_x^{\left(c\right)}=H^{+}{r}_y$

In formula, r _yfor all row in the covariance matrix of compression sampling signal being coupled together the vector formed, for all row in Cyclic Spectrum matrix being coupled together the vector formed; H ⁺for the MoorePenrose generalized inverse matrix of transformation matrix H; Wherein, H=Λ T ⁺, T ⁺for the Moore Penrose generalized inverse matrix of T, matrix $\mathrm{\&Lambda;}=Q_M(\mathrm{\&Phi;}\&CircleTimes;\mathrm{\&Phi;})/B_p,$ Φ is condensation matrix, matrix $T=(I\&CircleTimes;F)\underset{v=0}{\overset{N-1}{\mathrm{\&Sigma;}}}(D_v^T\&CircleTimes;G_v)/B_p,$ I is that N × N ties up unit matrix, and F is fast Fourier transform matrix, Q _m, B _p, G _vand D _vbe transformation matrix, for Kronecker amasss.

5. method according to claim 2, is characterized in that, the step of described real-time adjustment sampling rate specifically comprises:

Utilize the compression gains of detection probability determination compression sampling of described character rate, described modulation system and setting;

The compression ratio of compression sampling is obtained according to described compression gains, and by the sampling rate of described compression ratio and Nyquist sampling rate determination compression sampling.

6., based on a frequency spectrum detection device for compressed sensing, comprising:

Compression sampling unit, it carries out compression sampling based on selected sampling rate to the signal in spectrum environment and obtains compression sampling signal;

Spectral analysis unit, it utilizes described compression sampling signal recuperation Cyclic Spectrum;

Cavity detecting unit, whether it exists frequency spectrum cavity-pocket according in described Cyclic Spectrum estimated spectral environment.

7. device according to claim 6, it is characterized in that, also comprise pattern recognition unit and parameter adjustment unit, wherein, described pattern recognition unit identifies character rate and the modulation system of compression sampling signal based on described Cyclic Spectrum, and described parameter adjustment unit adjusts sampling rate in real time according to the detection probability of described modulation system, described character rate and setting.

8. device according to claim 7, is characterized in that, described pattern recognition unit is according to the character rate of following steps identification compression sampling signal and modulation system:

Cyclic Spectrum by compression sampling signal recuperation is projected to circulation frequency domain to obtain the circulation spectral envelope of compression sampling signal;

Carrier frequency and the character rate of compression sampling signal is extracted according to described circulation spectral envelope;

Utilize described carrier frequency and character rate to produce template signal, and described template signal is projected to circulation frequency domain to obtain the circulation spectral envelope of template signal;

The modulation system of template signal minimum for vector angle in the inner product of vectors of the circulation spectral envelope of compression sampling signal and the circulation spectral envelope of template signal is judged to be the modulation system of compression sampling signal.

9. the device according to any one of claim 6 to 8, is characterized in that, described spectral analysis unit recovers Cyclic Spectrum according to following formula:

$s_x^{\left(c\right)}=H^{+}{r}_y$

In formula, r _yfor all row in the covariance matrix of compression sampling signal being coupled together the vector formed, for all row in Cyclic Spectrum matrix being coupled together the vector formed; H ⁺for the MoorePenrose generalized inverse matrix of transformation matrix H; Wherein, H=Λ T ⁺, T ⁺for the Moore Penrose generalized inverse matrix of T, matrix $\mathrm{\&Lambda;}=Q_M(\mathrm{\&Phi;}\&CircleTimes;\mathrm{\&Phi;})/B_p,$ Φ is condensation matrix, matrix $T=(I\&CircleTimes;F)\underset{v=0}{\overset{N-1}{\mathrm{\&Sigma;}}}(D_v^T\&CircleTimes;G_v)/B_p,$ I is that N × N ties up unit matrix, and F is fast Fourier transform matrix, Q _m, B _p, G _vand D _vbe transformation matrix, for Kronecker amasss.

10. device according to claim 7, is characterized in that, described parameter adjustment unit adjusts sampling rate in real time according to following steps:

Utilize the compression gains of detection probability determination compression sampling of described character rate, described modulation system and setting;

The compression ratio of compression sampling is obtained according to described compression gains, and by the sampling rate of described compression ratio and Nyquist sampling rate determination compression sampling.