CN102394707B - Method for sensing broadband spectrum in modulation broadband converter sampling system - Google Patents

Abstract

The invention discloses a method for sensing a broadband spectrum in a modulation broadband converter sampling system. The method comprises the steps of support reconstruction, fine binary judgment, sub-band boundary estimation and combination. In the support reconstruction, coarse signal sub-bands are obtained. In the fine binary judgment, false-alarm sub-bands in the support reconstruction are eliminated. In the sub-band boundary estimation, a signal boundary in each sub-band is searched. In the combination, spectrum holes with over-small bandwidth are eliminated, and the adjacent spectrum holes in an over-small interval are combined. Compared with the broadband spectrum sensing method which is only dependent on the support reconstruction, the method provided by the invention has the advantage that: finer and more accurate spectrum hole information can be obtained.

Description

Broader frequency spectrum cognitive method in modulation wide-band transducer sampling system

Technical field

The present invention relates to wireless communication field, particularly the broader frequency spectrum cognitive method in modulation wide-band transducer sampling system.

Technical background

Because the mode that cognitive radio is the following user is used frequency spectrum, therefore before connection setup, it need to find the frequency spectrum hole not taken by primary user in wide-band, and this problem is broader frequency spectrum perception problems.In the prior art, broader frequency spectrum perception problems has multiple settling mode.For example, in a kind of mode, in AFE (analog front end), adopt a tunable narrow band filter, then by tuning narrow band filter, complete the searching and detecting to whole frequency range, but this mode can cause the perceived delay of growing.For improving search efficiency, in another kind of mode, can adopt narrow band filter group once to complete the analysis to whole perception frequency range, but this has increased hardware implementation complexity.Also having a kind of compromise proposal is to adopt wide-band high-speed AD converter (analog-to-digital converter, ADC) to sample to whole perception frequency range, then at numeric field, completes the detection judgement in frequency spectrum hole.This mode possesses very strong perception real-time, but because the sample frequency of ADC is limited, under Nyquist sampling theory, too high perception frequency range may exceed the sample frequency that current ADC can bear.

Along with the development of compressed sensing technology, researcher has proposed the method to analog signal sampling with the sampling rate lower than Nyquist speed, and these methods are called as sub-Nyquist (sub-Nyquist) method of sampling.The sub-Nyquist method of sampling having proposed at present mainly comprises periodically nonuniform sampling, Nyquist folding system, random demodulator sampling and modulation wide-band transducer sampling (modulated wideband converter, MWC).The people such as Mishali are at list of references 1 " Xampling:Analog to digital at sub-Nyquist rates; IET Circuits, Devices & Systems, vol.5, no.1, pp.8-20,2011 " in, compared in detail this few seed Nyquist method of samplings, from comparative result, can find out; than additive method, the sampling of modulation wide-band transducer have advantages of with actual signal mate, computation complexity is lower, realizability is high.

Fig. 1 is the theory diagram of modulation wide-band transducer.Specifically, signal x (t) inputs m passage simultaneously.At i passage, signal x (t) is T with the cycle _pmixing function multiply each other, by a cut-off frequency, be then 1/ (2T _s) low pass filter, filtered signal is with 1/T _sspeed sample, obtain sample sequence y _i[n].Due to 1/T _senough little, so existing commercial ADC can be used for sampling process.In addition, the sampling rate sum that an important feature of this sampling system is exactly m passage is still much smaller than Nyquist sample rate, i.e. mf _s< < f _nYQ(wherein, f _sfor the sample sample frequency of each passage of modulation wide-band transducer, f _nYQfor Nyquist sample rate), MWC is a seed Nyquist method of sampling.And the sampling rate of number of active lanes and single passage can be exchanged, this is highly beneficial to reducing hardware implementation complexity.

Analytical sampling sequences y now _i[n] and be input to the relation between the unknown signaling x (t) of modulation in wide-band transducer.Make f _p=1/T _p(f _pinverse for modulation wide-band transducer sampling mixing function cycle), f _s=1/T _s, by f _swith f _pcan be set as follows two interval: F _p=[f _p/ 2 ,+f _p/ 2], F _s=[f _s/ 2 ,+f _s/ 2].Consider i passage.Due to p _i(t) be periodic signal, so its Fourier expansion is:

$p_i\left(t\right)=\underset{l=-\&infin;}{\overset{\&infin;}{\mathrm{\&Sigma;}}}{c}_{\mathrm{il}}{e}^{\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="HSA00000589313600031.png,HSA00000589313600041.png"\; state="\{\{state\}\}">j</figure-callout>}\frac{2\mathrm{\&pi;}}{T_p}\mathrm{lt}}---\left(1\right)$

Wherein,

for MWC, there is following expression:

y(f)＝Az(f)，f∈F _s (2)

Wherein y (f) is m * 1 dimensional vector, and its i element is y _ithe discrete time Fourier conversion of [n], 1≤i≤m.Unknown matrix z (f)=[z ₁(f), z ₂(f) ..., z _l(f)] ^tlength is

represent to be more than or equal to a and with the immediate integer of a, wherein:

z _i(f)＝X(f+(i-L ₀-1)f _p)，1≤i≤L，f∈F _s (3)

The capable l column element of m * L matrix A i a _ilfor:

$a_{\mathrm{il}}=c_{i,{\mathrm{<figure-callout\; id="0"\; label="L"\; filenames="HSA00000589313600021.png,HSA00000589313600022.png"\; state="\{\{state\}\}">L</figure-callout>}}_0+1-l},$ 1≤i≤m，1≤l≤2L ₀+1 (4)

From formula (3), only require to obtain z (f), just can obtain original signal.Solving of z (f) depends on solving of equation (2), the in the situation that of known y (f) and A, solves z (f).In m * L matrix A, m < L, therefore, this problem is one and owes to determine problem.Under known z (f) is sparse situation, can adopt the signal reconfiguring method in compressive sensing theory to solve.The people such as Mishali have provided a kind of method for solving in list of references 2 " From theory to practice:Sub-Nyquist sampling of sparse wideband analog signals ", first according to low speed sample sequence vector y[n] structure matrix $Q={\&Integral;}_{f\&Element;F_s}y\left(f\right)y^H\left(f\right)\mathrm{df}=\underset{-\&infin;}{\overset{\&infin;}{\mathrm{\&Sigma;}}}y\left[n\right]y^T\left[n\right],$ Then with Q=VV ^hobtain matrix V, then by solving model V=AU, obtain the support S of U, due to support and the z (F of U _p) support equate, therefore obtain S=supp (z (F _p)); Finally, according to S, solve

and

in fact, support S corresponding to the subband that comprises signal, this feature will be applied in frequency spectrum perception.

Just because of the superiority of MWC, the people such as Mishali are at list of references 3 " Wideband spectrum sensing at sub-Nyquist Rates; IEEE Signal Process Mag, vol.28, no.4, pp.102-135,2011 " in, MWC is applied to cognitive radio wideband frequency spectrum perception, to perception frequency range, adopts MWC to sample lower than the sample rate of Nyquist speed, then by supporting reconstruct, find support; support corresponding to the frequency subband that comprises signal, thereby also obtained not comprising the frequency spectrum hole of signal.But this method is defectiveness also: because original signal is subject to noise pollution, the performance that supports restructing algorithm is limited, the support obtaining while supporting reconstruct may be false-alarm, even if support the support that reconstruct has obtained entirely accurate, support subband and also not necessarily by primary user's signal, taken (probably only capturing a part) completely.Therefore, the frequency spectrum sensing method that only depends on support reconstruct is also unreliable, and resulting frequency spectrum hole will be too coarse, and can omit important real frequency spectrum hole.

For improving performance, the present invention proposes and take a kind of cognitive radio wideband frequency spectrum cognitive method that MWC is the sub-Nyquist method of sampling, the method proposing is reconstructed into first step to support, after supporting reconstruct, also adopt other three steps to carry out explication de texte, these three steps are not comprise in the method that proposes of the people such as Mishali, are respectively: meticulous binary decision, subband border are estimated, merged.The cognitive method of reconstruct is supported in the only dependence proposing with respect to people such as Mishali, and the method that the present invention proposes can obtain meticulousr frequency spectrum hole information more accurately.

Summary of the invention

The object of the invention is to overcome existing method only depends on and supports the frequency spectrum sensing method of reconstruct unreliable, resulting frequency spectrum hole will be too coarse, and can omit the defect in important real frequency spectrum hole, thereby provide a kind of more accurately, meticulousr broader frequency spectrum cognitive method.

To achieve these goals, the invention provides a kind of broader frequency spectrum cognitive method of modulating in wide-band transducer sampling system, comprising:

Step 1), to doing support reconstruct by the formed original signal of resulting each sample sequence after modulation wide-band transducer sampling, obtain each subband label set S=supp (z (F that comprises narrow band signal in described original signal _p)), and obtain thus frequency spectrum hole V ₁; Wherein,

${\mathrm{<figure-callout\; id="1"\; label="V"\; filenames="HSA00000589313600041.png"\; state="\{\{state\}\}">V</figure-callout>}}_1=\underset{l\&NotElement;S}{\underset{1\&le;l\&le;2{\mathrm{<figure-callout\; id="0"\; label="L"\; filenames="HSA00000589313600021.png,HSA00000589313600022.png"\; state="\{\{state\}\}">L</figure-callout>}}_0+1}{\&cup;}}[(l-L_0-1)f_p-f_p/2,(l-L_0-1)f_p+f_p/2]$

Described f _pfor the inverse of modulation wide-band transducer sampling mixing function cycle,

f _sfor the sample sample frequency on each road of described modulation wide-band transducer, f _nYQfor Nyquist speed;

Step 2), according to step 1) resulting support S, calculate

wherein

representing matrix A _smoore-Penrose pseudoinverse, matrix A is the observing matrix of described modulation wide-band transducer sampling system, y[n] be the vector that each sample sequence of obtaining of sampling branch road forms, l ∈ S; Then build following binary test problems model, thereby do binary decision;

H ₀：z _l[n]＝w[n]

H ₁：z _l[n]＝s[n]+w[n]

S[n wherein] represent primary user's signal, w[n] represent white Gaussian noise, H ₀and H ₁represent respectively the judgement that primary user does not exist (being frequency spectrum hole) and primary user to exist;

To z _l[n] (l ∈ S) carries out after binary decision, if it is frequency spectrum hole, its corresponding index removes from support S; After all frequency spectrums hole has all removed from S, the new index set obtaining is designated as S '; According to described binary decision result, obtain extra frequency spectrum hole V ₂

${\mathrm{<figure-callout\; id="2"\; label="V"\; filenames="HSA00000589313600031.png,HSA00000589313600041.png"\; state="\{\{state\}\}">V</figure-callout>}}_2=\underset{l\&Element;S,l\&NotElement;S^{\&prime;}}{\&cup;}[(l-L_0-1)f_p-f_p/2,(l-L_0-1)f_p+f_p/2];$

Step 3), the frequency spectrum hole that step obtains is before merged.

In technique scheme, in described step 2) and step 3) between also comprise the step that subband border is estimated, this step comprises:

To z _lthe power spectral density of [n] (l ∈ S ') is carried out the estimation of subband border, finds out each narrow band signal of wherein comprising and the frequency boundary in frequency spectrum hole;

If

in there is P _lindividual frequency spectrum hole, lower boundary and the coboundary in p hole are respectively

with

obtain extra frequency spectrum hole

${\mathrm{<figure-callout\; id="3"\; label="V"\; filenames="HSA00000589313600041.png"\; state="\{\{state\}\}">V</figure-callout>}}_3=\underset{l\&Element;S^{\&prime;}}{\&cup;}\left\{\underset{p=1}{\overset{P_l}{\&cup;}}\right[(l-L_0-1)f_p+f_{z_l,\mathrm{low}}^{\left(p\right)},(l-L_0-1)f_p+f_{z_l,\mathrm{high}}^{\left(p\right)}\left]\right\}.$

In technique scheme, in described step 3) afterwards, also comprise step 3) operation that frequency spectrum hole after resulting merging is optimized, this operation comprises:

If the bandwidth in a certain frequency spectrum hole is less than Δ in the frequency spectrum hole set V after merging ₁, remove this frequency spectrum hole, if two adjacent frequency spectrum hole spacing are less than Δ in V ₂, judge that the subband between these two frequency spectrum holes is frequency spectrum hole, these two frequency spectrum holes are united two into one; Described Δ ₁and Δ ₂it is the parameters of cognitive radio system.

In technique scheme, described Δ ₁be set as the available minimal frequency of cognitive radio hole bandwidth, described Δ ₂be set as the possible minimum bandwidth of primary user's signal in cognitive radio working frequency range.

In technique scheme, in step 1) support restructuring procedure in, setting restructing algorithm iterations is its maximum iteration time that can move, thereby obtains the support element number of maximum possible.

In technique scheme, described step 2) binary decision in can adopt any blind binary detection method, comprises detection, the detection based on comentropy based on covariance matrix characteristic value.

In technique scheme, in the step of estimating on described subband border, adopt the method that many tap spectrums are estimated and the position finding method based on double threshold combines, comprising:

First adopt many tap spectrums to estimate z _lthe power spectral density of [n], obtains discrete power spectral density value ${\left\{{\hat{Z}}_l\right[n\left]\right\}}_{n=0}^{N_z-1};$

Follow basis

by two threshold parameter T are set ₁and T ₂, T wherein ₁> T ₂, carry out forward continuous average excision method, obtain two decision threshold T _land T _h, T wherein _l< T _h;

According to T _land T _h, by all T that are greater than _ladjacent PSD sampling point be classified as one bunch, if bunch in certain sample value be greater than T _h, judge that this bunch is as signal cluster; If

in there is not signal cluster, judge this z _l[n], corresponding to a broadband signal, whole subband is taken by signal, does not have frequency spectrum hole; If

in there is signal cluster, the border of signal cluster is border corresponding to frequency spectrum hole.

The present invention desirable following beneficial effect:

The present invention is by supporting analysis layer by layer and " peeling off " of reconstruct, meticulous binary decision, the estimation of subband border, obtain more and more careful frequency spectrum hole, finally at combining step, the frequency spectrum hole that each step obtains is above merged, and remove false frequency spectrum hole, the frequency spectrum hole that consolidation interval is narrow, thus final frequency spectrum hole obtained.Compare with the MWC Nyquist sampling broader frequency spectrum cognitive method that only relies on support reconstruct, the present invention can obtain more meticulous frequency spectrum hole information.

Accompanying drawing explanation

Fig. 1 is modulation wide-band transducer sampling block diagram.

Fig. 2 is broader frequency spectrum perception flow chart of the present invention.

The signal spectrum figure that Fig. 3 adopts for the emulation experiment done in one embodiment.

Fig. 4 is the frequency spectrum hole simulation result figure that signal obtains after supporting reconstruct in one embodiment.

Fig. 5 be in one embodiment signal through supporting the frequency spectrum hole simulation result figure after reconstruct and binary decision.

Fig. 6 is the simulation result figure of signal after subband border is estimated in one embodiment.

Fig. 7 is broader frequency spectrum perception simulation result figure in one embodiment.

Embodiment

Below in conjunction with drawings and Examples, the present invention is described in further detail.

If the frequency range of recognizing radio perception is [0, f _e], frequency range [0, f _e] in may comprise at most N primary user's signal (comprising negative frequency), primary user's signal bandwidth maximum possible is B, minimum may bandwidth be b, the available frequency spectrum of cognitive radio hole minimum bandwidth is B _c.Adopt MWC as shown in Figure 1 to sample to this frequency range, it is m>=2N that port number is set, and f is set _p>=B, arranges f _s>=f _p.After having sampled, obtain each low speed sample sequence y _i[n], 1≤i≤m.Resulting each low speed sample sequence y after sampling _i[n] (1≤i≤m), as original signal, as shown in Figure 2, can adopt following steps to carry out broader frequency spectrum perception in a most preferred embodiment:

(1) coarse support reconstruct

Coarse support reconstruct is for obtaining each subband label set S=supp (z (F that original signal comprises narrow band signal _p)).S=supp (z (F is supported _p)) after, just can obtain hole

${\mathrm{<figure-callout\; id="1"\; label="V"\; filenames="HSA00000589313600041.png"\; state="\{\{state\}\}">V</figure-callout>}}_1=\underset{l\&NotElement;S}{\underset{1\&le;l\&le;2{\mathrm{<figure-callout\; id="0"\; label="L"\; filenames="HSA00000589313600021.png,HSA00000589313600022.png"\; state="\{\{state\}\}">L</figure-callout>}}_0+1}{\&cup;}}[(l-L_0-1)f_p-f_p/2,(l-L_0-1)f_p+f_p/2]$

Wherein

The reconstructing method providing in the list of references 2 " From theory to practice:Sub-Nyquist sampling of sparse wideband analog signals " that coarse support reconstruct in this step can adopt " background technology " part to mention is realized, in restructuring procedure, can make the reinforce number obtaining by setting restructing algorithm iterations is 2N.

(2) meticulous binary decision

2N the son band that meticulous binary decision obtains for the coarse support reconstruct of step (1) carries out.The narrow band signal number comprising due to original signal may be less than N (comprising negative frequency), so z _l(f) (l ∈ S) may be still frequency spectrum hole.The support S obtaining according to the step in above-mentioned frequency spectrum sensing method (1), calculates wherein

representing matrix A _smoore-Penrose pseudoinverse, z _s[n] is by all z _lthe vector that [n] (l ∈ S) forms, z _l[n] is z _l(f) discrete time Fourier inverse transformation.The object of meticulous binary decision is judgement z _lwhether [n] (l ∈ S) is frequency spectrum hole, and this can be summed up as traditional binary test problems

H ₀：z _l[n]＝w[n]

H ₁：z _l[n]＝s[n]+w[n]

S[n wherein] represent primary user's signal, w[n] represent white Gaussian noise, H ₀and H ₁represent respectively the judgement that primary user does not exist (being frequency spectrum hole) and primary user to exist.This binary test problems can adopt existing any binary blind checking method to solve, and comprises detection based on characteristic value, the detection based on comentropy etc.

To z _l[n] (l ∈ S) carries out after binary decision, according to court verdict, can know whether it is frequency spectrum hole, if it is frequency spectrum hole, its corresponding index removes from support S.After all frequency spectrums hole has all removed from S, just can obtain new index set, be designated as S '.Through meticulous binary decision, obtain extra frequency spectrum hole

${\mathrm{<figure-callout\; id="2"\; label="V"\; filenames="HSA00000589313600031.png,HSA00000589313600041.png"\; state="\{\{state\}\}">V</figure-callout>}}_2=\underset{l\&Element;S,l\&NotElement;S^{\&prime;}}{\&cup;}[(l-L_0-1)f_p-f_p/2,(l-L_0-1)f_p+f_p/2]$

(3) subband border is estimated

Subband border is estimated for z _l[n] (l ∈ S ') carries out, and object is to find out z _lthe narrow band signal frequency boundary comprising in [n], thus frequency spectrum hole further obtained.

Adopted in the present embodiment many tap spectrums to estimate (multitaper method, MTM) the method antithetical phrase band border that the position finding method with based on double threshold (localization method based on double-thresholding, LAD) combines is estimated.

First adopt MTM to estimate z _lthe power spectral density of [n] (power spectral density, PSD).Calculate series of discrete Fourier conversion

$z_l^{\left(k\right)}\left(f\right)=\underset{n=0}{\overset{N_z-1}{\mathrm{\&Sigma;}}}{z}_l\left[n\right]v_n^{\left(k\right)}\exp (-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="HSA00000589313600031.png,HSA00000589313600041.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;fn}{T}_s)$

Wherein

splain sequence, k=0,1 ..., K ₀-1.

concentration of energy in resolution bandwidth 2W.Time bandwidth product provides the upper bound of number of taps,

mTM based on front several characteristic spectrums is:

${\hat{Z}}_l\left(f\right)=\left(\underset{k=0}{\overset{K_0-1}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_k{\left|z_l^{\left(k\right)}\left(f\right)\right|}^2\right){\left(\underset{k=0}{\overset{K_0-1}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_k\right)}^{-1}$

ρ wherein _kk characteristic spectrum characteristic of correspondence value.Getting f is centrifugal pump, obtains discrete PSD value ${\left\{{\hat{Z}}_l\right[n\left]\right\}}_{n=0}^{N_z-1}.$

Then according to discrete PSD value adopt LAD to carry out the estimation of subband border.LAD adopts forward continuous average excision method to calculate thresholding:

Step 1: right

carry out ascending order arrangement, the sequence after order is arranged is

calculate ξ=round (α N _z), obtain close to α N _zinteger value, 0 < α < 1 wherein.

Step 2: calculate $y_{\mathrm{\&xi;}}=\underset{n=1}{\overset{\mathrm{\&xi;}}{\mathrm{\&Sigma;}}}{\hat{Z}}_l^A\left[n\right]/\mathrm{\&xi;}.$

Step 3: if

wherein T is threshold parameter, makes ξ=ξ+1, jumps to step 2; Otherwise the process of thresholding is calculated in termination, obtained threshold value is Ty _ξ.

In LAD, by two threshold parameter T are set ₁and T ₂, T wherein ₁> T ₂, carry out forward continuous average excision method, obtain two decision threshold T _land T _h(T _l< T _h).According to T _land T _h, by all T that are greater than _ladjacent PSD sampling point be classified as one bunch, if bunch in certain sample value be greater than T _h, judge that this bunch is as signal cluster.If

in there is not signal cluster, judge this z _l[n], corresponding to a broadband signal, whole subband is taken by signal, does not have frequency spectrum hole.If

in there is signal cluster, the border of signal cluster is border corresponding to frequency spectrum hole.Order

in there is P _lindividual frequency spectrum hole, lower boundary and the coboundary in p hole are respectively

with

can obtain extra frequency spectrum hole

${\mathrm{<figure-callout\; id="3"\; label="V"\; filenames="HSA00000589313600041.png"\; state="\{\{state\}\}">V</figure-callout>}}_3=\underset{l\&Element;S^{\&prime;}}{\&cup;}\left\{\underset{p=1}{\overset{P_l}{\&cup;}}\right[(l-L_0-1)f_p+f_{z_l,\mathrm{low}}^{\left(p\right)},(l-L_0-1)f_p+f_{z_l,\mathrm{high}}^{\left(p\right)}\left]\right\}$

(4) merge

To three frequency spectrum hole V that step obtains above ₁, V ₂and V ₃merge, obtain

V＝V ₁∪V ₂∪V ₃

Due to the impact of various non-ideal factors, as implementation preferably, in the present embodiment, after frequency spectrum hole is merged, the V after being also combined analyzes for the last time, if the bandwidth in a certain frequency spectrum hole is less than B in V _c, remove this frequency spectrum hole, if two adjacent frequency spectrum hole spacing are less than b in V, judge that the subband between these two frequency spectrum holes is frequency spectrum hole, these two frequency spectrum holes are united two into one.But in other embodiments, also can save this analysis operation.

Thus, in the present embodiment, by analysis layer by layer and " peeling off " of first three step, obtain more and more careful frequency spectrum hole, finally in step 4, the frequency spectrum hole that each step obtains is above merged, and remove false frequency spectrum hole, thereby obtain final frequency spectrum hole.

Be more than the description to a most preferred embodiment of the present invention, in other preferred embodiment, also can realize broader frequency spectrum perception.In a preferred embodiment, described broader frequency spectrum cognitive method comprises the step of coarse support reconstruct, meticulous binary decision and merging, compare with the most preferred embodiment mentioned above, saved the step that subband border is estimated, also just there is no described frequency spectrum hole V ₃, corresponding, in combining step, when merging only to V ₁and V ₂merge.In another preferred embodiment, described broader frequency spectrum cognitive method comprises the step that coarse support reconstruct, subband border are estimated and merged.Compare with the most preferred embodiment mentioned above, saved the step of meticulous binary decision, also just there is no described frequency spectrum hole V ₂, corresponding, in combining step, when merging only to V ₁and V ₃merge.In addition, in optimum embodiment, it is the further operation to doing via resulting S ' after meticulous binary decision that subband border is estimated, in the present embodiment, owing to having saved the step of meticulous binary decision, therefore, subband border estimates it is the further operation to doing via the resulting support of coarse support reconstruct S.

Below adopt the method in aforementioned most preferred embodiment, by an emulation experiment, technique effect of the present invention is described further.

If: cognitive radio perception frequency range is [0,525] MHz, comprise altogether five narrow band signals, centre frequency is respectively 110MHz, 304MHz, 306MHz, 308MHz and 485MHz, the signal to noise ratio ratio of noise power (all narrow band signal power summations with) is 5dB, its power spectral density as shown in Figure 3, is noted because real signal power spectrum is symmetrical, so only drawn positive frequency part; The parameter of MWC system is: m=24, f _s=f _p=6MHz; Because 304MHz, 306MHz, tri-narrow band signals of 308MHz are crossed over altogether spectral bandwidth and are less than 6MHz, therefore, concerning MWC, they can regard a signal as, so signal number N=6.Thus, the total sample frequency of MWC is 144MHz, is far smaller than Nyquist sample rate 1.05GHz.

The iterations of choosing coarse support restructing algorithm is 12, has obtained a support S={7 who comprises 12 elements, and 11,29,37,69,70,106,107,139,147,165,169}, obtains thus ${\mathrm{<figure-callout\; id="1"\; label="V"\; filenames="HSA00000589313600041.png"\; state="\{\{state\}\}">V</figure-callout>}}_1=\underset{l\&NotElement;S}{\underset{1\&le;l\&le;175}{\&cup;}}[(l-88)6-3,(l-88)6+3].$ Fig. 4 has drawn V ₁, with low horizontal line, representing the frequency spectrum hole that judgement obtains, high horizontal line represents that court verdict is taken by primary user, gives the PSD of original signal in figure, to compare.In attention figure, only provided positive frequency part.As seen from the figure, have two (being four if comprise negative frequency) false-alarm subbands, these two false-alarm subbands will be removed in binary decision link.The meticulous binary decision of the present invention is for z _l[n] (l ∈ S) carries out.The binary blind checking method of selection based on minimax characteristic value ratio is to z _l[n] detects, and result is judged z ₁₁[n], z ₂₉[n], z ₁₄₇[n] and z ₁₆₅[n] is noise, S '={ 7,37,69,70,106,107,139, the 169} therefore obtaining.Fig. 5 has provided V ₁∪ V ₂result.Subband border is estimated for z _l[n] (l ∈ S ') carries out, and Fig. 6 has provided estimated result.As can be seen from the figure, the present invention has found signal subspace band border preferably.Fig. 7 is each frequency spectrum hole that broader frequency spectrum cognitive method of the present invention is finally determined.As can be seen from the figure, the present invention has accurately found the frequency spectrum hole comprising in original signal, and frequency spectrum hole resolution is far superior to f _p=6MHz, has better performance than the broader frequency spectrum cognitive method that only depends on support reconstruct.

Claims (6)

1. modulate the broader frequency spectrum cognitive method in wide-band transducer sampling system, comprising:

Step 1), to doing support reconstruct by the formed original signal of resulting each sample sequence after modulation wide-band transducer sampling, obtain each subband label set S=supp (z (F that comprises narrow band signal in described original signal _p)), F _p=[f _p/ 2 ,+f _p/ 2], and thus obtain frequency spectrum hole V ₁; Wherein,

Described f _pfor the inverse of modulation wide-band transducer sampling mixing function cycle,

f _sfor the sample sample frequency on each road of described modulation wide-band transducer, f _nYQfor Nyquist speed;

Step 2), according to step 1) resulting support S, calculate

wherein

representing matrix A _smoore-Penrose pseudoinverse, matrix A is the observing matrix of described modulation wide-band transducer sampling system, y[n] be the vector that each sample sequence of obtaining of sampling branch road forms, l ∈ S; Then build following binary test problems model, thereby do binary decision;

H ₀：z _l[n]=w[n]

H ₁：z _l[n]=s[n]+w[n]

S[n wherein] represent primary user's signal, w[n] represent white Gaussian noise, H ₀and H ₁represent respectively the judgement that primary user does not exist (being frequency spectrum hole) and primary user to exist;

To z _l[n] (l ∈ S) carries out after binary decision, if it is frequency spectrum hole, its corresponding index removes from support S; After all frequency spectrums hole has all removed from S, the new index set obtaining is designated as S '; According to described binary decision result, obtain extra frequency spectrum hole V ₂

Step 3), the frequency spectrum hole that step obtains is before merged;

In described step 2) and step 3) between also comprise the step that subband border is estimated, this step comprises:

To z _lthe power spectral density of [n] (l ∈ S ') is carried out the estimation of subband border, finds out each narrow band signal of wherein comprising and the frequency boundary in frequency spectrum hole;

If

in there is P _lindividual frequency spectrum hole, lower boundary and the coboundary in p hole are respectively with

obtain extra frequency spectrum hole

2. the broader frequency spectrum cognitive method in modulation wide-band transducer sampling system according to claim 1, it is characterized in that, in described step 3) afterwards, also comprise step 3) operation that frequency spectrum hole after resulting merging is optimized, this operation comprises:

If the bandwidth in a certain frequency spectrum hole is less than Δ in the frequency spectrum hole set V after merging ₁, remove this frequency spectrum hole, if two adjacent frequency spectrum hole spacing are less than Δ in V ₂, judge that the subband between these two frequency spectrum holes is frequency spectrum hole, these two frequency spectrum holes are united two into one; Described Δ ₁and Δ ₂it is the parameters of cognitive radio system.

3. the broader frequency spectrum cognitive method in modulation wide-band transducer sampling system according to claim 2, is characterized in that described Δ ₁be set as the available minimal frequency of cognitive radio hole bandwidth, described Δ ₂be set as the possible minimum bandwidth of primary user's signal in cognitive radio working frequency range.

4. the broader frequency spectrum cognitive method in modulation wide-band transducer sampling system according to claim 1 and 2, it is characterized in that, in step 1) support restructuring procedure in, setting restructing algorithm iterations is its maximum iteration time that can move, thereby obtains the support element number of maximum possible.

5. the broader frequency spectrum cognitive method in modulation wide-band transducer sampling system according to claim 1 and 2, it is characterized in that, described step 2) binary decision in can adopt any blind binary detection method, comprises detection, the detection based on comentropy based on covariance matrix characteristic value.

6. the broader frequency spectrum cognitive method in modulation wide-band transducer sampling system according to claim 1 and 2, it is characterized in that, in the step of estimating on described subband border, adopt the method that many tap spectrums are estimated and the position finding method based on double threshold combines, comprising:

First adopt many tap spectrums to estimate z _lthe power spectral density of [n], obtains discrete power spectral density value

Follow basis

by two threshold parameter T are set ₁and T ₂, T wherein ₁>T ₂, carry out forward continuous average excision method, obtain two decision threshold T _land T _h, T wherein _l<T _h;

Finally according to T _land T _h, by all T that are greater than _ladjacent PSD sampling point be classified as one bunch, if bunch in certain sample value be greater than T _h, judge that this bunch is as signal cluster; If

in there is not signal cluster, judge this z _l[n], corresponding to a broadband signal, whole subband is taken by signal, does not have frequency spectrum hole; If

in there is signal cluster, the border of signal cluster is border corresponding to frequency spectrum hole.

Images