CN101494508A - Frequency spectrum detection method based on characteristic cyclic frequency - Google Patents

Abstract

The invention relates to a frequency spectrum detection method based on characteristic cycle frequency, belonging to the technical field of wireless communication. When carrying out cycle spectrum processing of finite length to signals, the invention only carries out processing to cycle spectrum value at the characteristic cycle frequency of signals; simultaneously, the purpose of controlling detection probability and false-alarm probability of the signals is achieved by the adjustment of decision threshold. Based on the basic theory of cycle spectrum detection, the invention adopts finite signal length and selects the characteristic cycle frequency of signals to carry out frequency spectrum detection; when the cycle spectrum value at the characteristic cycle frequency is larger than preset threshold, the frequency spectrum is considered to be occupied or is considered to not be occupied. The proposal greatly improves the frequency spectrum detection performance and solves the problem that frequency detection is carried out under very low signal-to-noise ratio.

Description

Frequency spectrum detecting method based on characteristic cyclic frequency

Technical field

The present invention relates to a kind of detection method of wireless communication technology field, specifically, what relate to is a kind of frequency spectrum detecting method based on characteristic cyclic frequency.

Background technology

Cognitive radio technology for the primary user PU frequency range of Secondary Users' insertion authority provide may, whether wherein correctly detect some frequency range is one of core technology of cognitive radio by primary CU.Energy measuring and circulation spectrum detection algorithm are proposed in the frequency spectrum detection field.But in radio communication, the residing position of user may run into situations such as deep fade, " hiding website ", the received signal to noise ratio of Secondary Users is very low, energy measuring is greatly limited in this case, but circulation spectrum detection technique can obtain the quite good detecting performance at low signal-to-noise ratio, need very big amount of calculation yet on whole plane, do the computing of circulation spectrum, can't satisfy the demand of the real-time of frequency spectrum detection.Though frequency spectrum detection has had a variety of detection methods, in the present method that proposes, effectively do not solve the real-time detection problem under the low signal-to-noise ratio.

Find through retrieval prior art, existing detection scheme all not can solve the frequency spectrum detection problem under the low signal-to-noise ratio, as U.S. Patent number is " US 2008/0080604A1 ", Youngsik Hur, people such as Chang-HoLee in its patent " SPECTRUM-SENSING ALGORITHMS AND METHODS " (Chinese is " frequency spectrum detection algorithm and a method ") detection are divided into rough detection and examining surveyed for two steps, improve detection efficiency, but its detection method still is based on traditional energy measuring, does not fundamentally solve the deficiency of energy measuring itself.At application number is that " 200610020800.5 ", application people are the method that has proposed to give reliability to distributed node in University of Electronic Science and Technology, the patent of denomination of invention for China's invention of " a kind of distributed frequency spectrum detecting method based on reliability ", use the mode of Distributed Detection to improve the detection performance, but its individual node still is based on energy measuring, fails to solve energy measuring bigger deficiency affected by noise.Though the circulation of cyclo-stationary signal spectrum can be not affected by noise, many relevant papers have also been pointed out to circulate to compose the superiority incomparable with traditional detection method under low signal-to-noise ratio, but because therefore the complexity that the circulation spectrum is handled fails the circulation spectrum signature of signal is applied to actual system always.

Summary of the invention

The objective of the invention is at above shortcomings in the prior art, a kind of frequency spectrum detecting method based on characteristic cyclic frequency is provided, this scheme detects to handle between complexity and detection performance in the circulation spectrum compromises, promptly effectively improve the frequency spectrum detection performance, reduced the processing complexity that the circulation spectrum detects simultaneously.

The present invention is achieved by the following technical solutions, when the present invention carries out the circulation spectrum processing of finite length to signal, only the circulation spectrum numerical value at signal characteristic cycle frequency place handled; Simultaneously, reach the purpose of control signal detection probability and false alarm probability by adjustment to decision threshold.The present invention is based on the basic theories that the circulation spectrum detects, adopt time-limited signal length and select signal characteristic cycle frequency place to carry out frequency spectrum detection, for example, have characteristic point α=2f for amplitude modulation or phase-modulated signal _c, have characteristic point α=2f for FM signal _c± f _d, f wherein _cBe signal(-) carrier frequency, f _dBe the signal frequency difference.When the circulation at characteristic cyclic frequency place spectrum numerical value during, think that then frequency spectrum is occupied, otherwise think that frequency spectrum is unoccupied greater than predetermined threshold values thre.This scheme improves the performance of frequency spectrum detection greatly, has solved the problem of carrying out frequency spectrum detection under utmost point low signal-to-noise ratio.

The present invention includes following steps:

Step 1: system start-up, on checkout equipment, set the target frequency bands information that needs detection, i.e. carrier frequency f _c, and the sample frequency f of mould/number sample devices _sWith the false alarm probability P that satisfies the overall system performance requirement _f, passed through frequency translation as the target frequency bands information that needs detect, then carrier frequency is set to through the IF-FRE after the frequency translation, and is last, and the data length N that the circulation spectrum is handled is set;

Step 2: utilize the existed system information of target frequency bands, obtain the tabulation that this frequency range may modulation system, and table look-up and obtain the pairing characteristic frequency of these modulation systems, be designated as α ₀

Step 3: according to the data processing length of setting, the acquisition noise sample circulates to compose to handle and obtains noise characteristic circulation spectrum value sequence, according to noise circulation spectral sequence the generalized extreme value distribution parameter is estimated, obtain parameter κ, μ, the σ of generalized extreme value distribution (GEV), order estimates that the parameter that obtains is

Wherein κ is the form parameter of generalized extreme value distribution, and μ is a location parameter, and σ is a scale parameter.

The generalized extreme value distribution probability density function is as follows:

When κ ≠ 0

$f\left(x\right|k,\mathrm{\μ},\mathrm{\σ})=\left(\frac{1}{\mathrm{\σ}}\right)\exp (-{(1+k\frac{x-\mathrm{\μ}}{\mathrm{\σ}})}^{-\frac{1}{k}}){(1+k\frac{x-\mathrm{\μ}}{\mathrm{\σ}})}^{-1-\frac{1}{k}}---\left(1\right)$

$1+k\frac{(x-\mathrm{\μ})}{\mathrm{\σ}}>0$

When κ=0

$f\left(x\right|0,\mathrm{\μ},\mathrm{\σ})=\left(\frac{1}{\mathrm{\σ}}\right)\exp (-\exp (-\left(\frac{x-\mathrm{\μ}}{\mathrm{\σ}}\right))-\frac{x-\mathrm{\μ}}{\mathrm{\σ}})---\left(2\right)$

Step 4 is according to the false alarm probability value P that sets on checkout equipment _fThe parameter that obtains with step 3

Handle threshold value thre, it is as follows to handle expression formula:

Wherein

y _p＝-log(1-P _f)

Step 5, the signal of record after mould/number sample devices sampling is x (κ), N of choosing continuously in x (κ) sequence put the spectrum processing that circulates, but only the cycle of treatment spectral frequency is α ₀The value at place

And to get the circulation spectral frequency be α (f), ₀The maximum of place's circulation spectrum value is a feature circulation spectrum

Step 6 is declared α ₀The relation of place's feature circulation spectrum numerical value and thre, when ${\mathrm{<figure-callout\; id="0"\; label="C\; x\; x"\; filenames="A20091004669000111.png"\; state="\{\{state\}\}">C</figure-callout>}}_{x{x}_{}}^{{}_0}>\mathrm{thre}$ The time, think that target frequency bands has the signal transmission, when $C_x^{x_0}<\mathrm{thre}$ The time, think that target frequency bands does not have the signal transmission.

Among the present invention,, can obtain different judgement thresholdings, and, judge whether there is the signal transmission on the target frequency bands according to the circulation spectrum numerical value at characteristic cyclic frequency place and the magnitude relationship between this thresholding according to the demand of system to false alarm probability.

The present invention has effectively improved the frequency spectrum detection performance, has reduced the processing complexity that the circulation spectrum detects simultaneously.Detection method proposed by the invention is apparently higher than traditional energy measuring, at P _f=0.001 o'clock, the resulting detection probability of the method for energy measuring was almost nil, and detection probability of the present invention distributes in both cases and reaches 0.88 and 1; P _f=0.01 o'clock, the present invention's detection probability in both cases all reached 1, and the energy measuring probability is lower, is about 0.2 and 0.3.

Description of drawings

Fig. 1 is the branch spirogram of AM signal on the cycle frequency axle.

Fig. 2 the present invention draws the performance simulation curve chart;

Wherein: the performance of gained curve of the present invention and traditional energy measuring is relatively worked as p _f=10 ^-1Pairing detection probability is the testing result in the example of the present invention.

Embodiment

Below in conjunction with accompanying drawing embodiments of the invention are elaborated: present embodiment is being to implement under the prerequisite with the technical solution of the present invention, provided detailed execution mode and concrete operating process, but protection scope of the present invention is not limited to following embodiment.

Below be that embodiment further describes with the AM modulation signal: parameter setting does not influence generality.Adopt the AM signal to verify the correctness of detection method proposed by the invention as an example in this example.The AM signal is by white Gaussian noise channel AWGN, and frequency modulating signal is 10KHz, and carrier frequency is 1MHz, and the A/D sample frequency is 10MHz, and the sampling time is 0.4ms, and promptly used data length is N=4000, and the false alarm probability that system can receive is P _f=0.01.The selection characteristic frequency is α ₀=2f _c, as shown in Figure 1, at α ₀=2f _cThere is bigger non-zero-amplitude at the place.The concrete processing procedure of feature cycle detection is as follows:

1) the target frequency bands information that needs detection, i.e. carrier frequency f are set in system start-up on checkout equipment _c=1MHz, and the sample frequency f of mould/number sample devices _s=10MHz and overall system performance P _f=0.01, passed through frequency translation as the target frequency bands information that needs detect, then carrier frequency is set to through the IF-FRE after the frequency translation, and is last, and the data length N=2000 that the circulation spectrum is handled is set.

2), obtain the tabulation that this frequency range may modulation system, and table look-up and obtain the pairing characteristic frequency α of these modulation systems according to the existed system information of target frequency bands ₀=2f _c

3), gather N according to the data processing length of setting ²The spot noise sample circulates to compose to handle and obtains noise characteristic circulation spectrum value sequence, according to noise circulation spectral sequence the generalized extreme value distribution parameter is estimated, obtain parameter κ, μ, the σ of generalized extreme value distribution (GEV), wherein κ is the form parameter of generalized extreme value distribution, μ is a location parameter, σ is a scale parameter, uses maximal possibility estimation in this example and parameter is estimated detailed process is described below:

Under the situation of received signal not, own ship's noise sampled obtains N ²Spot noise is divided into the N section with the noise sample that obtains, every section sample noise sequence X _{N (i-1)-j}, i=1 ... N, j=1 ... N handles every section the noise sequence spectrum that circulates again, and processing procedure is as follows:

To i section sample noise, make x (j-1)=X _{N (i-1)-j}, j=1 ... N handles the circulation spectrum at its characteristic cyclic frequency place, and process is as follows:

$X\left[v\right]=\underset{k=0}{\overset{N-1}{\mathrm{\&Sigma;}}}x\left(k\right)e^{-\mathrm{<figure-callout\; id="2"\; label="i"\; filenames="A20091004669000111.png"\; state="\{\{state\}\}">i</figure-callout>}2\mathrm{\&pi;vk}/N}---\left(4\right)$

$S^{{\mathrm{\&alpha;}}_0}\left(f_p\right)=\frac{1}{(N-1)T_s}\frac{1}{L}\underset{v=-\frac{L-1}{2}}{\overset{v=\frac{L-1}{2}}{\mathrm{\&Sigma;}}}X(f_p+\frac{{\mathrm{\&alpha;}}_0}{2F_s}+v)X*(f_p-\frac{{\mathrm{\&alpha;}}_0}{2F_s}+v)W\left(v\right)---\left(5\right)$

$C_i^{{\mathrm{\&alpha;}}_0}=\underset{f_p}{\max }\left\{S^{\mathrm{\&alpha;}}\left(f_p\right)\right\}---\left(6\right)$

F wherein _sBe sample frequency, W is a M point window function, M=7 in this example, and window function is got Hanning window.

After handling, the circulation spectrum obtains N=4000 noise characteristic circulation spectrum value sequence

I=1 ... N is according to the above feature noise circulation spectral sequence that obtains

I=1 ... N carries out maximal possibility estimation to generalized extreme value distribution (GEV), order $y_i=C_i^{{\mathrm{\&alpha;}}_0},$ When κ ≠ 0, the max log natural function that GEV distributes is:

$l(k,\mathrm{\&mu;},\mathrm{\&sigma;})=-N\log \mathrm{\&sigma;}-(1+1/k)\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\log [1+k\left(\frac{y_i-\mathrm{\&mu;}}{\mathrm{\&sigma;}}\right)]-\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{[1+k\left(\frac{y_i-\mathrm{\&mu;}}{\mathrm{\&sigma;}}\right)]}^{-\frac{1}{k}}---\left(7\right)$

Here requirement

$1+k\left(\frac{y_i-\mathrm{\&mu;}}{\mathrm{\&sigma;}}\right)>0$

Otherwise the likelihood function value is zero, and corresponding log-likelihood function value is-∞ that the parameter of the white Gaussian noise behind the circulation spectral transformation satisfies this condition.Following formula (7) is maximized about parameter (κ, μ, σ), even $\frac{\&PartialD;l}{\&PartialD;k}=0,$ $\frac{\&PartialD;l}{\&PartialD;\mathrm{\&mu;}}=0,$ $\frac{\&PartialD;l}{\&PartialD;\mathrm{\&delta;}}=0$ Can obtain the maximum likelihood estimation that GEV distributes.In actual treatment and since κ when being positioned near zero (by actual sample

Estimate that resulting κ is positioned near zero), the expression formula of natural logrithm likelihood function usefulness κ=0 o'clock, as follows:

$l(\mathrm{\&mu;},\mathrm{\&sigma;})=-N\log \mathrm{\&sigma;}-\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\left(\frac{y_i-\mathrm{\&mu;}}{\mathrm{\&sigma;}}\right)-\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\exp \{-\left(\frac{y_i-\mathrm{\&mu;}}{\mathrm{\&sigma;}}\right)\}---\left(8\right)$

Order $\frac{\&PartialD;l}{\&PartialD;\mathrm{\&mu;}}=0,$ $\frac{\&PartialD;l}{\&PartialD;\mathrm{\&delta;}}=0$ Thereby obtain the likelihood equation group, abbreviation can get:

$\left\{\begin{array}{c}\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{e}^{-(y_i-\widehat{\mathrm{\&mu;}})/\widehat{\mathrm{\&sigma;}}}=N\\ \underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}(y_i-\widehat{\mathrm{\&mu;}})(1-e^{-(y_i-\widehat{\mathrm{\&mu;}})/\widehat{\mathrm{\&sigma;}}})=N\widehat{\mathrm{\&sigma;}}\end{array}\right.---\left(9\right)$

The likelihood equation group does not have explicit solution, finds the solution by numerical method, obtains the estimated value of parameter μ, σ

Again formula (7) is maximized about κ $\frac{\&PartialD;l}{\&PartialD;k}=0$ Obtain the estimated value of parameter κ

Thereby obtain the estimated value of parameter κ, μ, σ $\hat{k}=-0.0213,$ $\widehat{\mathrm{\&mu;}}=0.5508,$ $\widehat{\mathrm{\&sigma;}}=3.7195;$

4) according to the false alarm probability value P that on checkout equipment, sets _fThe parameter that obtains with step 3 Handle thresholding thre=6.1429;

5) signal of record after mould/number sample devices sampling is x (κ), N of choosing continuously in x (κ) sequence put the spectrum processing that circulates, but only the cycle of treatment spectral frequency is α ₀The value at place

And to get the circulation spectral frequency be α (f), ₀The maximum of place's circulation spectrum value is a feature circulation spectrum

Concrete processing formula is as follows:

$X\left[v\right]=\underset{k=0}{\overset{N-1}{\mathrm{\&Sigma;}}}x\left(k\right)e^{-\mathrm{<figure-callout\; id="2"\; label="i"\; filenames="A20091004669000111.png"\; state="\{\{state\}\}">i</figure-callout>}2\mathrm{\&pi;vk}/N}---\left(10\right)$

$S^{{\mathrm{\&alpha;}}_0}\left(f_p\right)=\frac{1}{(N-1)T_s}\frac{1}{L}\underset{v=-\frac{L-1}{2}}{\overset{v=\frac{L-1}{2}}{\mathrm{\&Sigma;}}}X(f_p+\frac{{\mathrm{\&alpha;}}_0}{2F_s}+v)X*(f_p-\frac{{\mathrm{\&alpha;}}_0}{2F_s}+v)W\left(v\right)---\left(11\right)$

$C_i^{{\mathrm{\&alpha;}}_0}=\underset{f_p}{\max }\left\{S^{\mathrm{\&alpha;}}\left(f_p\right)\right\}---\left(12\right)$

6) judge α ₀Place's feature circulation spectrum numerical value

With the relation of thre, when ${\mathrm{<figure-callout\; id="0"\; label="C\; x\; x"\; filenames="A20091004669000111.png"\; state="\{\{state\}\}">C</figure-callout>}}_{x{x}_{}}^{{}_0}>\mathrm{thre}$ The time, think that target frequency bands has the signal transmission, when $C_x^{x_0}<\mathrm{thre}$ The time, think that target frequency bands does not have the signal transmission.

As shown in Figure 2, provided SNR in the drawings to be respectively-21dbh and-testing result of two kinds of situations of 17db, as can be seen from the figure, detection method proposed by the invention is apparently higher than traditional energy measuring, at P _f=0.001 o'clock, the resulting detection probability of the method for energy measuring was almost nil, and detection probability of the present invention distributes in both cases and reaches 0.88 and 1; P _f=0.01 o'clock, the present invention's detection probability in both cases all reached 1, and the energy measuring probability is lower, is about 0.2 and 0.3.

From implementation process of the present invention as can be seen, this method is based on characteristic cyclic frequency, counting of the cycle frequency of required processing has only a bit or some, and circulation spectrum detect need to calculate on the whole circulation frequency axis have a few, if counting, signal sampling is N, the point that circulation spectrum so detects required calculating is the N point, and amount of calculation is counted with calculating and is directly proportional, so the present invention has effectively reduced the complexity of coming out that detects based on the circulation spectrum.

Claims (3)

1, a kind of frequency spectrum detecting method based on circulation spectrum signature value is characterized in that may further comprise the steps:

Step 1: system start-up, on checkout equipment, set the target frequency bands information that needs detection, i.e. carrier frequency f _c, and the sample frequency f of mould/number sample devices _sWith the false alarm probability P that satisfies the overall system performance requirement _f, passed through frequency translation as the target frequency bands information that needs detect, then carrier frequency is set to through the IF-FRE after the frequency translation, and is last, and the data length N that the circulation spectrum is handled is set;

Step 2: according to the existed system information of target frequency bands, obtain the tabulation that this frequency range may modulation system, and table look-up and obtain the pairing characteristic frequency of these modulation systems, be designated as α ₀

Step 3: according to the data processing length of setting, the acquisition noise sample circulates to compose to handle and obtains noise characteristic circulation spectrum value sequence, according to noise circulation spectral sequence the generalized extreme value distribution parameter is estimated, obtained parameter k, μ, the σ of generalized extreme value distribution, order estimates that the parameter that obtains is Wherein k is the form parameter of generalized extreme value distribution, and μ is a location parameter, and σ is a scale parameter;

Step 4: the parameter that obtains according to the false alarm probability value Pf that on checkout equipment, sets and step 3

Handle threshold value thre;

Step 5: the signal of record after mould/number sample devices sampling is x (k), N of choosing continuously in x (k) sequence put the spectrum processing that circulates, but only the cycle of treatment spectral frequency is α ₀The value at place

And to get the circulation spectral frequency be α ₀The maximum of place's circulation spectrum value is a feature circulation spectrum

Step 6: judge α ₀The relation of place's feature circulation spectrum numerical value and thre, when $C_x^{{\mathrm{\&alpha;}}_0}>\mathrm{thre}$ The time, think that target frequency bands has the signal transmission, when $C_x^{{\mathrm{\&alpha;}}_0}<\mathrm{thre}$ The time, think that target frequency bands does not have the signal transmission.

2, the frequency spectrum detecting method based on circulation spectrum signature value according to claim 1 is characterized in that, in the step 3, the probability density function of generalized extreme value distribution is as follows:

When k ≠ 0

$f\left(x\right|k,\mathrm{\&mu;},\mathrm{\&sigma;})=\left(\frac{1}{\mathrm{\&sigma;}}\right)\exp (-{(1+k\frac{x-\mathrm{\&mu;}}{\mathrm{\&sigma;}})}^{-\frac{1}{k}}){(1+k\frac{x-\mathrm{\&mu;}}{\mathrm{\&sigma;}})}^{-1-\frac{1}{k}}$

$1+k\frac{(x-\mathrm{\&mu;})}{\mathrm{\&sigma;}}>0$

When k=0

$f\left(x\right|0,\mathrm{\&mu;},\mathrm{\&sigma;})=\left(\frac{1}{\mathrm{\&sigma;}}\right)\exp (-\exp (-\left(\frac{x-\mathrm{\&mu;}}{\mathrm{\&sigma;}}\right))-\frac{x-\mathrm{\&mu;}}{\mathrm{\&sigma;}}).$

3, the frequency spectrum detecting method based on circulation spectrum signature value according to claim 1 is characterized in that, in the step 4, the processing expression formula of handling threshold value thre is as follows:

Wherein: y _p=-log (1-P _f).

Images