CN104780008A - Broadband spectrum sensing method based on self-adaptive compressed sensing - Google Patents

Abstract

The invention relates to the technical field of radio communication, in particular to a broadband spectrum sensing method based on self-adaptive compressed sensing. The broadband spectrum sensing method comprises the following steps: a sensing time slot is divided, that is, the sensing time slot is divided into a fixed sensing sub-time slot and a plurality of self-adaptive sensing sub-time slots; a frequency range is divided into a plurality of broad frequency ranges, and a selectable broad frequency range is selected in a self-adaptive manner as a frequency range to be sensed each time according to communication requirements and known spectrum resource conditions; compressed sampling is carried out on the fixed sensing sub-time slot, an observation vector on the fixed sensing sub-time slot is obtained, and primary signal spectrum reconstruction is carried out; compressed sampling is carried out on each self-adaptive sensing sub-time slot, the observation vector is updated, and spectrum reconstruction is carried out based on the result of the previous sensing sub-time slot; whether convergence conditions are satisfied or not is judged, compressed sampling is continued on the next self-adaptive sensing sub-time slot if the convergence conditions are not satisfied, and the self-adaptive sensing process is ended if the convergence conditions are satisfied; the spectrum sensing result is finally output, and residual self-adaptive sensing sub-time slot resources can still be used for normal communication.

Description

A kind of broader frequency spectrum cognitive method based on self-adapting compressing perception

Technical field

The present invention relates to radio communication technology field, be specifically related to a kind of broader frequency spectrum cognitive method based on self-adapting compressing perception.

Background technology

In prior art, traditional frequency spectrum sensing method comprises energy measuring perception, matched filtering perception and cyclostationary characteristic perception, and the frequency spectrum sensing method introduced in the scientific and technical literature of China and outside China and patent documentation belongs to above-mentioned a few class.Along with the extensive use of broadband wireless network, wireless communication system works in the frequency range of more Gao Gengkuan, and awareness apparatus must have broader frequency spectrum detectability, and bandwidth may be high to GHz.Conventional method does not possess the perception in broader frequency spectrum scope, and measurement bandwidth is wider, required sampling rate and processing speed higher, hard-wired technical difficulty and cost are also larger, limit the application of frequency spectrum perception technology.In recent years, people are based on the intrinsic openness consideration of broadband wireless signal frequency spectrum, compression sampling theory (Compressed Sensing Theory) thought is incorporated into frequency spectrum perception field, carry out lack sampling with the speed far below Nyquist rate to signal, recycling restructing algorithm recovers primary signal.Z.Tian has delivered " Compressed Sensing for Wideband Cognitive Radios " on international conference ICASSP 2007, adopts compressed sensing technology to carry out broader frequency spectrum perception first; Xian Electronics Science and Technology University proposes the patent of invention of " the compression frequency spectrum sensing method based on autocorrelation matrix reconstitution ", utilizes autocorrelation matrix characteristic to realize frequency spectrum perception that is quick, low complex degree.Compression frequency spectrum sensing method efficiently solves the problem being difficult to carry out perception on broader frequency spectrum, but still have that perception efficiency is not high, strong to environmental suitability, perceived spectral scope and perceived resolution, the conflicting shortcoming of algorithm complex.

For solving perception efficiency and the problem to environmental suitability, a kind of two-step compression frequency spectrum sensing method with self adaptation perception thought is proposed in the article " A two-step compressed spectrum sensing scheme for wideband cognitive radios " that Y.Wang delivers in 2010IEEE Globecom meeting, the first step estimates the spectrum sparse degree of received spectrum signal, the sampling point quantity of second step needed for spectrum sparse degree adjustment signal reconstruction, realizes adaptive compression frequency spectrum perception.Number of patent application 201310544294.X, denomination of invention is that " a kind of degree of rarefication self-adapting compressing frequency spectrum sensing method based on progressive step-length " proposes similar viewpoint, and difference is to utilize closed-loop structure according to degree of rarefication judgement adjustment observation statistic.Said method has adaptive characteristic, can improve perception efficiency to a certain extent, have the adaptability to environment spectrum sparse degree, but two kinds of methods all need additionally to estimate to carry out frequency spectrum perception again to spectrum sparse degree; In addition these class methods precision that spectrum sparse degree is estimated and threshold value more responsive, in actual applications not directly, difficulty is comparatively large, reliability is not high.

Future broadband wireless communication systems will hold increasing user, the service of more high-performance, more contents is provided, traffic carrying capacity constantly increases the contradiction with radio spectrum resources anxiety, in order to solve the problem that frequency spectrum resource is short and the availability of frequency spectrum is low, there has been proposed the concept of cognitive radio (CR, Cognitive Radio).Cognitive radio, under the prerequisite ensureing authorized user proper communication, allows unauthorized user dynamic access to distribute to authorized user but not by the idle frequency spectrum that it uses, thus improves system spectrum utilance.Unaffected to authorizing the use of frequency range in order to ensure authorized user, cognitive user must quickly and accurately to authorizing frequency range carry out perception, detects which frequency range occupied or be in the free time.Therefore, frequency spectrum perception is technology the most key in cognitive radio and top priority.

Summary of the invention

In order to overcome defect of the prior art, provide a kind of broader frequency spectrum cognitive method based on self-adapting compressing perception, in real time according to frequency spectrum perception reconstruction result termination perception initiatively, the broader frequency spectrum occupancy that better adaptation is different and received signal to noise ratio situation, save frequency spectrum perception required time, improve device data efficiency of transmission, the method have employed broadband software radio framework and self adaptation perception Frequency Band Selection simultaneously, while guarantee ultrabroad band sensing range, frequency spectrum perception result is meticulousr, accurate, and method realizes more simple.

The present invention is achieved through the following technical solutions: a kind of broader frequency spectrum cognitive method based on self-adapting compressing perception, said method comprising the steps of:

Step 1: divide perception time slot, carries out frequency spectrum perception to the perception time slot that the cycle occurs, is a fixing perception sub-slots and several self adaptation perception sub-slots by described perception time-slot division;

Step 2: frequency range is divided into some wide-bands, confirm to treat the optional wide-band of perception, the wide frequency range to gigahertz (GHZ) is divided into several optional frequency ranges, each described optional frequency range still occupies the broadband of tens or hundreds of megahertz, according to communication requirement and known frequency spectrum resource situation, each only adaptive selection optional frequency range is as treating perception frequency range, by radio-frequency head mixing and filtering process, treat that the broadband analog signal of perception frequency range is moved low frequency or intermediate frequency and separates, for follow-up perception analysis by described;

Step 3: compression sampling on fixing perception sub-slots, fixes on perception sub-slots and treats that perception frequency band signals carries out compression sampling to adaptively selected described in step 1, obtains the observation vector on fixing perception sub-slots;

Step 4: the signal spectrum reconstruct observation vector fixed described in step 3 on perception sub-slots being carried out to particle swarm optimization algorithm;

Step 5: compression sampling on described self adaptation perception sub-slots, upgrades observation vector, carries out the frequency spectrum reconfiguration based on conjugate gradient algorithms;

Step 6: judge whether the frequency spectrum reconfiguration result on this self adaptation perception sub-slots meets the condition of convergence, does not meet and then returns the compressed sensing that step 5 carries out next self adaptation perception sub-slots, satisfied then enter step 7;

Step 7: export final frequency spectrum perception result, utilizes residue self adaptation perception sub-slots, " free time " frequency sub-band carries out proper communication.

Further, in described method, step 4 comprises the following steps:

1) according to the observation vector y on fixing perception sub-slots ₀reconstructed spectrum can be converted to and ask l ₁the optimal solution of-norm Least optimization problem, shown in (1):

$\underset{\tilde{r}}{\min }{\left|\right|\tilde{r}\left|\right|}_1+\mathrm{\λ}{\left|\right|y_0-{\tilde{A}}_0\tilde{r}\left|\right|}_2^2\·\·\·\·\·\·\left(1\right),$

Wherein λ is positive weights coefficient, and the another kind of expression formula of formula (1) is:

$\underset{\tilde{r},r={[\mathrm{Re}{\left(\tilde{r}\right)}^T,\mathrm{Im}{\left(\tilde{r}\right)}^T]}^T}{\min }\mathrm{\Ω}(\tilde{r},r),\mathrm{\Ω}(\tilde{r},r)={\left|\right|\tilde{r}\left|\right|}_1+\mathrm{\λ}{\left|\right|y_0^{\′}-A_{0}r\left|\right|}_2^2\·\·\·\·\·\·\left(2\right),$

for the target function of PSO algorithm, $y_0^{\′}={[{\left(y_0\right)}^T,0_{M_0\×1}^T]}^T,$ $A_0=\left[\begin{array}{cc}\mathrm{Re}\left({\tilde{A}}_0\right)& -\mathrm{Im}\left({\tilde{A}}_0\right)\\ \mathrm{Im}\left({\tilde{A}}_0\right)& \mathrm{Re}\left({\tilde{A}}_0\right)\end{array}\right];$

2) initialization particle swarm optimization algorithm parameter, makes particle swarm optimization algorithm iteration variable be k, maximum iteration time K _max, population population N _p, inertial factor ω and Studying factors are respectively γ ₁, γ ₂;

3) k=0 is made, at solution space initialization N _pindividual particle, the position vector and the velocity vector that define its 2N dimension are respectively $z_i^{\left(0\right)}(i=\mathrm{1,2},...,N_p)$ And $v_i^{\left(0\right)}(i=\mathrm{1,2},...,N_p);$

4) make $r^{\left(k\right)}=z_i^{\left(k\right)},$ ${\tilde{r}}^{\left(k\right)}={[z_i^{\left(k\right)}\left(0\right),...,z_i^{\left(k\right)}(N-1)]}^T+j{[z_i^{\left(k\right)}\left(N\right),...,z_i^{\left(k\right)}(2N-1)]}^T,$ By N _pindividual position vector substitutes into formula (2) calculating target function value respectively relatively find particle self history optimal solution vector the history optimal solution vector of acquisition is compared with particles all in population $p_{\mathrm{all}}^{\left(k\right)}=\underset{z_i^{\left(k^{\′}\right)},k^{\′}=\mathrm{0,1},...,k,i=\mathrm{1,2},...,N_p}{\min }{\mathrm{\Ω}}_i^{\left(k^{\′}\right)};$

5) velocity vector and the position vector of population is upgraded according to the optimal solution vector of current iteration:

$\left\{\begin{array}{c}{v}_i^{(k+1)}={\mathrm{\ωv}}_i^{\left(k\right)}+{\mathrm{\λ}}_1(p_i^{\left(k\right)}-z_i^{\left(k\right)})+{\mathrm{\λ}}_2(p_{\mathrm{all}}^{\left(k\right)}-z_i^{\left(k\right)})\\ z_i^{(k+1)}=z_i^{\left(k\right)}+v_i^{(k+1)}\end{array}\right.\·\·\·\·\·\·\left(3\right),$

Parameter ω, γ ₁, γ ₂according to the number that embody rule selects [0,1] interval;

6) k=k+1, when meeting k=K _maxtime stop particle swarm optimization algorithm iteration, the solution vector on fixing perception sub-slots ${\hat{r}}_0=p_{\mathrm{all}}^{(K_{\max }-1)},$ Frequency spectrum reconfiguration result is ${\stackrel{\widehat{~}}{r}}_0={[{\hat{r}}_0\left(0\right),...,{\hat{r}}_0(N-1)]}^T+j{[{\hat{r}}_0\left(N\right),...,{\hat{r}}_0(2N-1)]}^T,$ Satisfied then enter described step 5, do not meet k=K _maxtime, continue iteration with initialization particle swarm optimization algorithm, turn back to the 4th in described step 4) step.

Further, in described method, step 5 comprises the following steps:

1) first initiation parameter l=1 is set;

2) on l self adaptation perception sub-slots, carry out compression sampling, self adaptation perception sub-slots width is T _a, obtain one group of T _af _sthe vector of individual sampled value will the observation vector y of compressed sensing process is used for l-1 sub-slots _l-1merge, obtain the M on l self adaptation perception sub-slots _l=(T ₀+ lT _a) f _sdimension observation vector observation vector w _lfor white Gaussian noise vector, for the M upgraded along with self adaptation perception sub-slots _l× N ties up observing matrix,

Similarly, optimization problem is converted to a kind of representation:

$\underset{\tilde{r},r={[\mathrm{Re}{\left(\tilde{r}\right)}^T,\mathrm{Im}{\left(\tilde{r}\right)}^T]}^T}{\min }\mathrm{\Ψ}(\tilde{r},r),\mathrm{\Ψ}(\tilde{r},r)={\left|\right|\tilde{r}\left|\right|}_1+\mathrm{\λ}{\left|\right|y_l^{\′}-A_{l}r\left|\right|}_2^2\·\·\·\left(4\right),$ Wherein $y_l^{\′}={[{\left(y_l\right)}^T,0_{M_l\×1}^T]}^T,$

$A_l=\left[\begin{array}{cc}\mathrm{Re}\left({\tilde{A}}_l\right)& -\mathrm{Im}\left({\tilde{A}}_l\right)\\ \mathrm{Im}\left({\tilde{A}}_l\right)& \mathrm{Re}\left({\tilde{A}}_l\right)\end{array}\right],$ Y _l' and A _lcan be easy at y ' _l-1and A _l-1basis upgrades acquisition;

3) set conjugate gradient algorithms iteration variable as t, maximum iteration time T _{l, max}, α _tit is step factor;

4) t=0 is made, the initial value of solution

5) conjugate gradient algorithms are adopted to calculate:

$r_l^{(t+1)}=r_l^{\left(t\right)}+{\mathrm{\α}}_t{d}_l^{\left(t\right)}\·\·\·\·\·\·\left(5\right),$

$d_l^{\left(t\right)}=\left\{\begin{array}{cc}-g_l^{\left(t\right)},& t=0\\ -g_l^{\left(t\right)}+{\mathrm{\β}}_t{d}_l^{(t-1)},& t>0\end{array}\right.\·\·\·\·\·\·\left(6\right),$

Wherein $g_l^{\left(t\right)}=\▿\mathrm{\Ψ}\left(r_l^{\left(t\right)}\right),$ Parameter ${\mathrm{\β}}_t=g_l^{T\left(t\right)}(g_l^{\left(t\right)}-g_l^{(t-1)})/{\left|\right|g_l^{(t-1)}\left|\right|}^2;$

6) t=t+1, when meeting t=T _{l, max}time stop conjugate gradient algorithms iteration, the solution vector on l self adaptation perception sub-slots ${\hat{r}}_l={\hat{r}}_l^{T_{l,\max }},$ Frequency spectrum reconfiguration result is ${\stackrel{\widehat{~}}{r}}_l={[{\hat{r}}_l\left(0\right),...,{\hat{r}}_l(N-1)]}^T+j{[{\hat{r}}_l\left(N\right),...,{\hat{r}}_l(2N-1)]}^T,$ Then enter step 6; Do not meet t=T _{l, max}time, conjugate gradient algorithms continue iteration, turn back to the described 5th) step.

Further, the step 6 of described method comprises the following steps:

First judge whether to satisfy condition minimum tolerable error when ε is optimal solution convergence, if satisfy condition, stops perception, enters step 7; Not satisfying condition, then l=l+1, turn back to the 2nd in described step 5) step carries out on next self adaptation perception sub-slots compressed sensing.

Further, described step 7 comprises the following steps:

1) the final frequency spectrum reconfiguration value obtained is exported calculate the performance number treating each sub-band signals within the scope of perceived frequency $p\left(n\right)={\left|\stackrel{\widehat{~}}{r}\left(n\right)\right|}^2,n=\mathrm{0,1},..,N-1;$

2) judge which frequency sub-band is idle or occupied by threshold method; Residue self adaptation perception sub-slots can be utilized, idle frequency sub-band carries out proper communication.

Further, analog signal information transducer is used to carry out compression sampling in the step 3 of described method.

Compared with prior art, superior effect is:

1. the inventive method adopts compression frequency spectrum sensing method directly can carry out accurate real-time perception to the broader frequency spectrum of hundred MHz, utilizes software radio framework and self adaptation Frequency Band Selection strategy, can realize the wide broadband perception to GHz frequency range.

2. the inventive method has the feature of adaptive spectrum perception, can according to spectrum sparse degree different in actual application environment and received signal to noise ratio situation, adaptive adjustment detecting period, reduces awareness apparatus power consumption, improves user data transmission efficiency and throughput.Under some environment that spectrum sparse degree is lower or received signal to noise ratio is higher, compared with the method adopting fixing time slot perception with tradition, the present invention can shorten the detecting period of 20% ~ 70%.

3. present invention reduces the requirement to high-speed AD converter, relax the restriction to follow-up signal processing speed, can need to be configured relevant parameter according to application, take into account the index such as perceived bandwidth, perceived spectral resolution, flexible design, simple and be easy to realize.

Accompanying drawing explanation

Fig. 1 is the broader frequency spectrum cognitive method flow chart of steps based on self-adapting compressing perception of the present invention;

Fig. 2 is the broader frequency spectrum cognitive method intermediate frequency spectrum perception broadband spectrum division schematic diagram based on self-adapting compressing perception of the present invention;

Fig. 3 is of the present invention based on perception time-slot division schematic diagram in the broader frequency spectrum cognitive method of self-adapting compressing perception;

Fig. 4 is of the present invention based on radio frequency front end in the broader frequency spectrum cognitive method of self-adapting compressing perception and analog signal information converter structure block diagram;

Fig. 5 is the original broader frequency spectrum signal schematic representation of the broader frequency spectrum cognitive method middle width strip spectrum signal reconstruction result based on self-adapting compressing perception of the present invention;

Fig. 6 is the broader frequency spectrum signal schematic representation of the reconstruct of the broader frequency spectrum cognitive method middle width strip spectrum signal reconstruction result based on self-adapting compressing perception of the present invention;

Fig. 7 be of the present invention based on different compression sampling in the broader frequency spectrum cognitive method of self-adapting compressing perception than the detection probability schematic diagram in situation when false alarm probability is 1%;

Fig. 8 is of the present invention based on the frequency spectrum perception Performance comparision schematic diagram in the broader frequency spectrum cognitive method of self-adapting compressing perception under different frequency bands occupancy and state of signal-to-noise.

Reference numeral is as follows:

In figure, the direction of arrow is the method for the invention steps flow chart direction.

Embodiment

Below in conjunction with accompanying drawing, the specific embodiment of the invention is described in further detail.

As shown in Figure 1, illustrate a kind of broader frequency spectrum cognitive method based on self-adapting compressing perception provided by the invention, said method comprising the steps of:

Step 1: the perception time slot that awareness apparatus occurred in the cycle carries out frequency spectrum perception, definition perception time slot is made up of a fixing perception sub-slots and several self adaptation perception sub-slots;

Step 2: the wide frequency range to gigahertz (GHZ) (GHz) will may be divided into several optional frequency ranges, each optional frequency range still occupies the broadband of tens or hundreds of megahertz (MHz), according to communication requirement and known frequency spectrum resource situation, awareness apparatus only needs adaptive selection optional frequency range as treating perception frequency range at every turn, by radio-frequency head mixing, filtering process, by treating that the broadband analog signal of perception frequency range is moved Low Medium Frequency and separates, for follow-up perception analysis;

Step 3: on fixing perception sub-slots, adaptively selected is treated that perception frequency band signals carries out compression sampling, with analog signal information transducer (Analog-to-Information Converter, AIC) realize compression sampling, obtain observation vector y on the fixing perception sub-slots of observation ₀;

Analog signal information transducer (AIC) is made up of high speed PRBS waveform generator, multiplier, integrate and dump filter, low speed analog to digital converter, the pseudo random sequence waveform that pseudo random sequence waveform generator produces the cycle is multiplied with analog signal, output waveform is accumulated by integrate and dump filter, with far below nyquist sampling rate f _nthe speed f of > 2W _streat perceptual signal and carry out lack sampling, wherein W is the highest frequency treating perception frequency band signals moving Low Medium Frequency, usually, makes f _n=α f _s, α is compression sampling ratio; Fixing perception sub-slots time width is T ₀, compression sampling value composition M ₀=T ₀f _sdimension observation vector wherein be treat that the N of perception frequency range ties up spectral vectors, represent and will treat that perception band limits evenly divides the spectrum value of N number of sub-band; w ₀be that to meet average be 0, standard variance is δ ²the white Gaussian noise of distribution; m ₀× N ties up observing matrix, that N ties up inverse discrete Fourier transformer inverse-discrete matrix, S ₀=D ₀c is the compression sampling matrix set up on fixing perception sub-slots, and C is pseudo random sequence matrix:

Pseudo random sequence [c ₀, c ₁..., c _l-1] be with f _nto the discrete representation of the pseudo random sequence waveform sampling of one-period, D ₀filtering sampling matrix:

Matrix characterizes integrate and dump filter form, compression sampling speed f _sand fixing perception slot time width T ₀;

Step 4: to the observation vector y that fixing perception sub-slots obtains ₀carry out the frequency spectrum reconfiguration based on particle swarm optimization algorithm (PSO):

1) according to the observation vector y on fixing perception sub-slots ₀reconstructed spectrum can be converted to and ask l ₁the optimal solution of-norm Least optimization problem, shown in (1),

wherein λ is positive weights coefficient, and a kind of expression formula of formula (1) is: $\underset{\tilde{r},r={[\mathrm{Re}{\left(\tilde{r}\right)}^T,\mathrm{Im}{\left(\tilde{r}\right)}^T]}^T}{\min }\mathrm{\Ω}(\tilde{r},r),\mathrm{\Ω}(\tilde{r},r)={\left|\right|\tilde{r}\left|\right|}_1+\mathrm{\λ}{\left|\right|y_0^{\′}-A_{0}r\left|\right|}_2^2\·\·\·\·\·\·\left(2\right),$ for the target function of (PSO) algorithm, $y_0^{\′}={[{\left(y_0\right)}^T,0_{M_0\×1}^T]}^T,$ $A_0=\left[\begin{array}{cc}\mathrm{Re}\left({\tilde{A}}_0\right)& -\mathrm{Im}\left({\tilde{A}}_0\right)\\ \mathrm{Im}\left({\tilde{A}}_0\right)& \mathrm{Re}\left({\tilde{A}}_0\right)\end{array}\right];$

2) initialization particle swarm optimization algorithm (PSO) parameter, makes particle swarm optimization algorithm (PSO) iteration variable be k, maximum iteration time K _max, population population N _p, inertial factor ω and Studying factors are respectively γ ₁, γ ₂;

3) k=0 is made, at solution space initialization N _pindividual particle, the position vector and the velocity vector that define its 2N dimension are respectively $z_i^{\left(0\right)}(i=\mathrm{1,2},...,N_p)$ And $v_i^{\left(0\right)}(i=\mathrm{1,2},...,N_p);$

4) make $r^{\left(k\right)}=z_i^{\left(k\right)},$ ${\tilde{r}}^{\left(k\right)}={[z_i^{\left(k\right)}\left(0\right),...,z_i^{\left(k\right)}(N-1)]}^T+j{[z_i^{\left(k\right)}\left(N\right),...,z_i^{\left(k\right)}(2N-1)]}^T,$ By N _pindividual position vector substitutes into formula (2) calculating target function value respectively relatively find particle self history optimal solution vector the history optimal solution vector of acquisition is compared with particles all in population $p_{\mathrm{all}}^{\left(k\right)}=\underset{z_i^{\left(k^{\′}\right)},k^{\′}=\mathrm{0,1},...,k,i=\mathrm{1,2},...,N_p}{\min }{\mathrm{\Ω}}_i^{\left(k^{\′}\right)};$

5) velocity vector and the position vector of population is upgraded according to the optimal solution vector of current iteration:

$\left\{\begin{array}{c}{v}_i^{(k+1)}={\mathrm{\ωv}}_i^{\left(k\right)}+{\mathrm{\λ}}_1(p_i^{\left(k\right)}-z_i^{\left(k\right)})+{\mathrm{\λ}}_2(p_{\mathrm{all}}^{\left(k\right)}-z_i^{\left(k\right)})\\ z_i^{(k+1)}=z_i^{\left(k\right)}+v_i^{(k+1)}\end{array}\right.\·\·\·\·\·\·\left(3\right),$ Parameter ω, γ ₁, γ ₂according to the number that embody rule selects [0,1] interval;

6) k=k+1, when meeting k=K _maxtime stop PSO algorithm iteration, the solution vector on fixing perception sub-slots ${\hat{r}}_0=p_{\mathrm{all}}^{(K_{\max }-1)},$ Frequency spectrum reconfiguration result is ${\stackrel{\widehat{~}}{r}}_0={[{\hat{r}}_0\left(0\right),...,{\hat{r}}_0(N-1)]}^T+j{[{\hat{r}}_0\left(N\right),...,{\hat{r}}_0(2N-1)]}^T,$ Enter step 5; Do not meet k=K _maxtime, particle swarm optimization algorithm (PSO) continues iteration, gets back to above-mentioned steps 4);

Step 5: continue to treat perception frequency band signals and carry out compression sampling on self adaptation perception sub-slots, upgrades observation vector, carries out the frequency spectrum reconfiguration based on conjugate gradient algorithms (CG):

1) initiation parameter l=1;

2) on l self adaptation perception sub-slots, carry out compression sampling, self adaptation perception sub-slots width is T _a, obtain one group of T _af _sthe vector of individual sampled value will the observation vector y of compressed sensing process is used for l-1 sub-slots _l-1merge, obtain the M on l self adaptation perception sub-slots _l=(T ₀+ lT _a) f _sdimension observation vector observation vector w _lfor white Gaussian noise vector, for the M upgraded along with self adaptation perception sub-slots _l× N ties up observing matrix,

similarly, optimization problem is converted to a kind of representation:

$\underset{\tilde{r},r={[\mathrm{Re}{\left(\tilde{r}\right)}^T,\mathrm{Im}{\left(\tilde{r}\right)}^T]}^T}{\min }\mathrm{\Ψ}(\tilde{r},r),\mathrm{\Ψ}(\tilde{r},r)={\left|\right|\tilde{r}\left|\right|}_1+\mathrm{\λ}{\left|\right|y_l^{\′}-A_{l}r\left|\right|}_2^2\·\·\·\·\·\·\left(4\right),$

Wherein $y_l^{\′}={[{\left(y_l\right)}^T,0_{M_l\×1}^T]}^T,$ $A_l=\left[\begin{array}{cc}\mathrm{Re}\left({\tilde{A}}_l\right)& -\mathrm{Im}\left({\tilde{A}}_l\right)\\ \mathrm{Im}\left({\tilde{A}}_l\right)& \mathrm{Re}\left({\tilde{A}}_l\right)\end{array}\right],$ Y _l' and A _lcan be easy at y ' _l-1and A _l-1basis upgrades acquisition;

3) set conjugate gradient method (CG) iteration variable as t, maximum iteration time T _{l, max}, α _tit is step factor;

4) t=0 is made, the initial value of solution

5) conjugate gradient method (CG) is adopted to calculate:

$r_l^{(t+1)}=r_l^{\left(t\right)}+{\mathrm{\α}}_t{d}_l^{\left(t\right)}\·\·\·\·\·\·\left(5\right),$

$d_l^{\left(t\right)}=\left\{\begin{array}{cc}-g_l^{\left(t\right)},& t=0\\ -g_l^{\left(t\right)}+{\mathrm{\β}}_t{d}_l^{(t-1)},& t>0\end{array}\right.\·\·\·\·\·\·\left(6\right),$

Wherein $g_l^{\left(t\right)}=\▿\mathrm{\Ψ}\left(r_l^{\left(t\right)}\right),$ Parameter ${\mathrm{\β}}_t=g_l^{T\left(t\right)}(g_l^{\left(t\right)}-g_l^{(t-1)})/{\left|\right|g_l^{(t-1)}\left|\right|}^2;$

6) t=t+1, when meeting t=T _{l, max}time stop conjugate gradient method (CG) iteration, the solution vector on l self adaptation perception sub-slots frequency spectrum reconfiguration result is:

${\stackrel{\widehat{~}}{r}}_l={[{\hat{r}}_l\left(0\right),...,{\hat{r}}_l(N-1)]}^T+j{[{\hat{r}}_l\left(N\right),...,{\hat{r}}_l(2N-1)]}^T,$ Enter step 6;

Do not meet t=T _{l, max}time, conjugate gradient method (CG) algorithm continues iteration, gets back to the 5th of step 5 the) step;

Step 6: judge whether to satisfy condition minimum tolerable error when ε is optimal solution convergence.If satisfy condition, stop perception, enter step 7; Not satisfying condition, then l=l+1, get back to the 2nd of step 5 the) step carries out on next self adaptation perception sub-slots compressed sensing;

Step 7: export the final frequency spectrum reconfiguration value that this method obtains calculate the performance number treating each sub-band signals within the scope of perceived frequency judge which frequency sub-band is idle or occupied by threshold method, remaining self adaptation perception sub-slots still can be used as communication time slot and is used.

As shown in Figure 2, wherein awareness apparatus can allow to access f ₀~ f _ifrequency range, the wideband frequency range that the present embodiment provides is f ₀=400MHz, f _i=2.4GHz, whole frequency range is divided into several optional frequency ranges, and each optional frequency range is wide is 250MHz, and optional frequency range is divided into again N number of frequency sub-band.Black cylinder in Fig. 2 represents that corresponding frequency sub-band exists primary user's signal, all the other white portions represent that corresponding frequency sub-band is unoccupied, awareness apparatus requires to detect " free time " frequency sub-band in real time, carry out perception to the time cycle of awareness apparatus beyond transfer of data, perception time slot width is 4 μ s, and the time slot for high speed data transfer is also 4 μ s, in this case, the perception Time Slot Occupancy resource of system 50%, reduces data transmission efficiency.As shown in Figure 3, by perception time-slot division be 1 T ₀the fixing perception sub-slots of=1 μ s and 12 T _athe self adaptation perception sub-slots of=0.25 μ s, awareness apparatus is Received signal strength in the perception time slot in cycle.Software radio radio-frequency front-end as shown in Figure 4 and analog signal information converter structure block diagram, by amplification, mixing, filtering process, treat that perception frequency range broadband signal receives by interested, recycling analog signal information transducer carries out compression sampling; Fixing perception sub-slots completes compression sampling, obtains fixing perception sub-slots observation vector y ₀, adopt particle swarm optimization algorithm to obtain the solution vector of perception on stator time slot and frequency spectrum reconfiguration result and the initial value using the sensing results on stator time slot as follow-up perception; Self adaptation perception sub-slots one by one completes compression sampling, upgrades the observation vector y on self adaptation perception sub-slots _l, adopt conjugate gradient algorithms to solve reconstruction signal whether judgement meets the frequency spectrum reconfiguration result condition of convergence do not meet and then continue compressed sensing on next self adaptation perception sub-slots, satisfied then obtain final frequency spectrum reconfiguration result the optional frequency range power spectrum of perception is treated in calculating, determines that rapidly which continuous print frequency sub-band can allow access, if this optional frequency range spectral occupancy is high, so can select other optional frequency ranges when next perception.Once the convergence of frequency spectrum reconfiguration value in the present invention, stop frequency spectrum perception process at once, residue still not can be used for transfer of data for the time slot of frequency spectrum perception, the highest shorten 70% detecting period, data transmission resources utilance is promoted to 70%-85% by 50%.In order to check the performance of the inventive method further, we from frequency spectrum perception result, parameter on the impact of perceptual performance, carried out a large amount of emulation experiment to aspects such as the adaptability of spectrum signal degree of rarefication and received signal to noise ratio.As shown in Figure 5 and Figure 6, it is the result adopting the present invention to carry out signal reconstruction at broad frequency range, shown in Fig. 5 is the broader frequency spectrum of certain the optional frequency range treating perception, shown in Fig. 6 is the broader frequency spectrum obtained this optional frequency range perception and signal reconstruction, utilize the present invention can reconstruct at the 2nd self adaptation perception sub-slots settling signal, remaining 10 self adaptation perception sub-slots all can be used for transfer of data, improves system communication efficiency.As shown in Figure 7, the present invention than the detection probability in situation when false alarm probability is 1%, investigates detection probability when norator frequency range is occupied or idle in optional frequency range at different compression sampling.Can find out, even if frequency spectrum detection probability of the present invention is still higher under lower signal to noise ratio, when compression sampling is than being α=1/16, when signal to noise ratio is 5dB, when ensureing that false alarm probability is 1%, detection probability is 54%.If increase compression sampling ratio further, its frequency spectrum detection performance has remarkable lifting, but when compression sampling ratio is increased to a certain degree, as compression sampling is increased to α=1/4 than by α=1/8, perceptual performance improvement is also not obvious, but the sampling rate of corresponding ADC adding at double, therefore, selects moderate compression sampling can obtain the optimization compromise between perceptual performance and implementation complexity than parameter.As shown in Figure 8, the frequency spectrum perception Performance comparision of the present invention under different frequency bands occupancy and state of signal-to-noise, mean square error (MSE) value between the reconstruction signal utilizing frequency spectrum perception to recover and primary signal carrys out the frequency spectrum detection performance of assessment algorithm, mean square error (MSE) value less expression sensing results is more accurate, and then performance is poorer on the contrary.The square mean error amount of frequency spectrum perception result under different frequency bands occupancy when solid line in Fig. 8 represents that the received signal to noise ratio of perception user is 5dB; The square mean error amount of frequency spectrum perception result under different frequency bands occupancy when dotted line represents that perception user received signal to noise ratio is 0dB, when the frequency band occupancy of the broader frequency spectrum signal treating perception is lower, namely when signal is more sparse, the present invention has signal reconstruction convergence rate faster, only need less perception time slot to complete perception, and remaining perception time slot can be preserved for carrying out transfer of data; In addition, when each user's Received signal strength average signal-to-noise ratio is higher, more easily frequency spectrum perception is completed fast.Therefore, the present invention can adapt to frequency band occupancies different in practical application and state of signal-to-noise well, and adaptive adjustment detecting period, can reduce equipment operating power consumption from perception user perspective, improves system transfers efficiency.

The present invention is not limited to above-mentioned execution mode, and when not deviating from flesh and blood of the present invention, any distortion that it may occur to persons skilled in the art that, improvement, replacement all fall into protection scope of the present invention.

Claims (6)

1., based on a broader frequency spectrum cognitive method for self-adapting compressing perception, it is characterized in that, said method comprising the steps of:

Step 1: divide perception time slot, carries out frequency spectrum perception to the perception time slot that the cycle occurs, is a fixing perception sub-slots and several self adaptation perception sub-slots by described perception time-slot division;

Step 2: frequency range is divided into some wide-bands, confirm to treat the optional wide-band of perception, the wide frequency range to gigahertz (GHZ) is divided into several optional frequency ranges, each described optional frequency range still occupies the broadband of tens or hundreds of megahertz, according to communication requirement and known frequency spectrum resource situation, each only adaptive selection optional frequency range is as treating perception frequency range, by radio-frequency head mixing and filtering process, treat that the broadband analog signal of perception frequency range is moved low frequency or intermediate frequency and separates, for follow-up perception analysis by described;

Step 3: compression sampling on fixing perception sub-slots, fixes on perception sub-slots and treats that perception frequency band signals carries out compression sampling to adaptively selected described in step 1, obtains the observation vector on fixing perception sub-slots;

Step 4: the signal spectrum reconstruct observation vector fixed described in step 3 on perception sub-slots being carried out to particle swarm optimization algorithm;

Step 5: compression sampling on described self adaptation perception sub-slots, upgrades observation vector, carries out the frequency spectrum reconfiguration based on conjugate gradient algorithms;

Step 6: judge whether the frequency spectrum reconfiguration result on described self adaptation perception sub-slots meets the condition of convergence, does not meet and then returns the compressed sensing that step 5 carries out next self adaptation perception sub-slots, satisfied then enter step 7;

Step 7: export final frequency spectrum perception result, utilizes residue self adaptation perception sub-slots, " free time " frequency sub-band carries out proper communication.

2., according to claim 1 based on the broader frequency spectrum cognitive method of self-adapting compressing perception, it is characterized in that, in described method, step 4 comprises the following steps:

1) according to the observation vector y on fixing perception sub-slots ₀reconstructed spectrum can be converted to and ask l ₁the optimal solution of-norm Least optimization problem, shown in (1):

$\underset{\tilde{r}}{\min }{\left|\right|\widetilde{r\left|\right|}}_1+\mathrm{\λ}{\left|\right|y_0-{\tilde{A}}_0\tilde{r}\left|\right|}_2^2\·\·\·\·\·\·\left(1\right),$

Wherein λ is positive weights coefficient, and the another kind of expression formula of formula (1) is:

$\underset{\tilde{r},r={[\mathrm{Re}{\left(\tilde{r}\right)}^T,\mathrm{Im}{\left(\tilde{r}\right)}^T]}^T}{\min }\mathrm{\Ω}(\tilde{r},r),\mathrm{\Ω}(\tilde{r},r)={\left|\right|\tilde{r}\left|\right|}_1+\mathrm{\λ}{\left|\right|y_0^{\′}-A_{0}r\left|\right|}_2^2\·\·\·\·\·\·\left(2\right),$

for the target function of PSO algorithm, $y_0^{\′}={[{\left(y_0\right)}^T,0_{M_0\×1}^T]}^T,$ $A_0=\left[\begin{array}{cc}\mathrm{Re}\left({\tilde{A}}_0\right)& -\mathrm{Im}\left({\tilde{A}}_0\right)\\ \mathrm{Im}\left({\tilde{A}}_0\right)& \mathrm{Re}\left({\tilde{A}}_0\right)\end{array}\right];$

2) initialization particle swarm optimization algorithm parameter, makes particle swarm optimization algorithm iteration variable be k, maximum iteration time K _max, population population N _p, inertial factor ω and Studying factors are respectively γ ₁, γ ₂;

3) k=0 is made, at solution space initialization N _pindividual particle, the position vector and the velocity vector that define its 2N dimension are respectively $z_i^{\left(0\right)}(i=\mathrm{1,2},...,N_p)$ And $v_i^{\left(0\right)}(i=\mathrm{1,2},...,N_p);$

4) make $r^{\left(k\right)}=z_i^{\left(k\right)},$ ${\tilde{r}}^{\left(k\right)}={[z_i^{\left(k\right)}\left(0\right),...,z_i^{\left(k\right)}(N-1)]}^T+j{[z_i^{\left(k\right)}\left(N\right),...,z_i^{\left(k\right)}(2N-1)]}^T,$ By N _pindividual position vector substitutes into formula (2) calculating target function value respectively relatively find particle self history optimal solution vector the history optimal solution vector of acquisition is compared with particles all in population $p_{\mathrm{all}}^{\left(k\right)}=\underset{z_i^{\left(k^{\′}\right)},k^{\′}=\mathrm{0,1},...,k,i=\mathrm{1,2},...,N_p}{\min }{\mathrm{\Ω}}_i^{\left(k^{\′}\right)};$

5) velocity vector and the position vector of population is upgraded according to the optimal solution vector of current iteration:

$\left\{\begin{array}{c}{v}_i^{(k+1)}=\mathrm{\ω}{v}_i^{\left(k\right)}+{\mathrm{\λ}}_1(p_i^{\left(k\right)}-z_i^{\left(k\right)})+{\mathrm{\λ}}_2(p_{\mathrm{all}}^{\left(k\right)}-z_i^{\left(k\right)})\\ z_i^{(k+1)}=z_i^{\left(k\right)}+v_i^{(k+1)}\end{array}\right.......\left(3\right)$ Parameter ω, γ ₁, γ ₂according to the number that embody rule selects [0,1] interval;

6) k=k+1, when meeting k=K _maxtime stop particle swarm optimization algorithm iteration, the solution vector on fixing perception sub-slots ${\hat{r}}_0=p_{\mathrm{all}}^{(K_{\max }-1)},$ Frequency spectrum reconfiguration result is ${\hat{\tilde{r}}}_0={[{\hat{r}}_0\left(0\right),...,{\hat{r}}_0(N-1)]}^T+j{[{\hat{r}}_0\left(N\right),...,{\hat{r}}_0(2N-1)]}^T,$ Satisfied then enter described step 5, do not meet k=K _maxtime, continue iteration with initialization particle swarm optimization algorithm, turn back to the 4th in described step 4) step.

3., according to claim 1 based on the broader frequency spectrum cognitive method of self-adapting compressing perception, it is characterized in that, in described method, step 5 comprises the following steps:

1) first initiation parameter l=1 is set;

2) on l self adaptation perception sub-slots, carry out compression sampling, self adaptation perception sub-slots width is T _a, obtain one group of T _af _sthe vector of individual sampled value will the observation vector y of compressed sensing process is used for l-1 sub-slots _l-1merge, obtain the M on l self adaptation perception sub-slots _l=(T ₀+ lT _a) f _sdimension observation vector observation vector w _lfor white Gaussian noise vector, for the M upgraded along with self adaptation perception sub-slots _l× N ties up observing matrix,

similarly, optimization problem is converted to one

Plant representation:

$\underset{\tilde{r},r={[\mathrm{Re}{\left(\tilde{r}\right)}^T,\mathrm{Im}{\left(\tilde{r}\right)}^T]}^T}{\min }\mathrm{\ψ}(\tilde{r},r),\mathrm{\ψ}(\tilde{r},r)={\left|\right|\tilde{r}\left|\right|}_1+\mathrm{\λ}{\left|\right|y_l^{\′}-A_{l}r\left|\right|}_2^2\·\·\·\·\·\·\left(4\right),$ Wherein $y_l^{\′}={[{\left(y_l\right)}^T,0_{M_l\×1}^T]}^T,$

$A_l=\left[\begin{array}{cc}\mathrm{Re}\left({\tilde{A}}_l\right)& -\mathrm{Im}\left({\tilde{A}}_l\right)\\ \mathrm{Im}\left({\tilde{A}}_l\right)& \mathrm{Re}\left({\tilde{A}}_l\right)\end{array}\right],$ Y ' _land A _lcan be easy at y ' _l-1and A _l-1basis upgrades acquisition;

3) set conjugate gradient algorithm iteration variable as t, maximum iteration time T _{l, max}, α _tit is step factor;

4) t=0 is made, the initial value of solution

5) conjugate gradient algorithms are adopted to calculate:

$r_l^{(t+1)}=r_l^{\left(t\right)}+{\mathrm{\α}}_t{d}_l^{\left(t\right)}......\left(5\right),$

$d_l^{\left(t\right)}=\left\{\begin{array}{cc}-g_l^{\left(t\right)},& t=0\\ -g_l^{\left(t\right)}+{\mathrm{\β}}_t{d}_l^{(t-1)},& t>0\end{array}\right.......\left(6\right),$

Wherein $g_l^{\left(t\right)}=\▿\mathrm{\ψ}\left(r_l^{\left(t\right)}\right),$ Parameter ${\mathrm{\β}}_t=g_l^{T\left(t\right)}(g_l^{\left(t\right)}-g_l^{(t-1)})/{\left|\right|g_l^{(t-1)}\left|\right|}^2;$

6) t=t+1, when meeting t=T _{l, max}time stop CG iteration, the solution vector on l self adaptation perception sub-slots ${\hat{r}}_l={\hat{r}}_l^{T_{l,\max }},$ Frequency spectrum reconfiguration result is ${\hat{\tilde{r}}}_l={[{\hat{r}}_l\left(0\right),...,{\hat{r}}_l(N-1)]}^T+j{[{\hat{r}}_l\left(N\right),...,{\hat{r}}_l(2N-1)]}^T,$ Then enter step described rapid 6; Do not meet time, conjugate gradient algorithms continue iteration, turn back to the described 5th) step.

4., according to claim 1 based on the broader frequency spectrum cognitive method of self-adapting compressing perception, it is characterized in that, the step 6 of described method comprises the following steps:

First judge whether to satisfy condition minimum tolerable error when ε is optimal solution convergence, if satisfy condition, stops perception, enters step 7; Not satisfying condition, then l=l+1, turn back to the 2nd in described step 5) step carries out on next self adaptation perception sub-slots compressed sensing.

5., according to claim 1 based on the broader frequency spectrum cognitive method of self-adapting compressing perception, it is characterized in that, described step 7 comprises the following steps:

1) the final frequency spectrum reconfiguration value obtained is exported calculate the performance number treating each sub-band signals within the scope of perceived frequency $p\left(n\right)={\left|\hat{\tilde{r}}\left(n\right)\right|}^2,n=\mathrm{0,1},...,N-1;$

2) judge which frequency sub-band is idle or occupied by threshold method; Utilize residue self adaptation perception sub-slots, idle frequency sub-band carries out proper communication.

6. according to claim 1 based on the broader frequency spectrum cognitive method of self-adapting compressing perception, it is characterized in that, in the step 3 of described method, use analog signal information transducer to carry out compression sampling.