CN102546061A - Self-adaptive time-frequency hole detection method based on wavelet transformation - Google Patents

Abstract

The invention discloses a self-adaptive time-frequency hole detection method based on wavelet transformation. A wavelet transformation technology is adopted to self-adaptively carry out time frequency domain subsection on time continuous signals or discrete-time signals and judge whether each time frequency resource block is idle or not. The method comprises the following steps of: (1) firstly carrying out time domain segmentation on cognition user monitoring and receiving signals to ensure the occupied frequency band of each main user to be constant in the segment; (2) carrying out Fourier transformation on receiving signals of each sub time interval in the existing method to obtain the signal frequency spectrum of each sub time interval, and carrying out wavelet transformation to the frequency spectrum to detect the border of each frequency range so as to realize frequency range division; and (3) synthesizing self-adaptive time and frequency domain segmentation results in the last two steps to comprehensively analyze each time frequency resource block and judge whether the time frequency resource block is idle or not. The method disclosed by the invention realizes time domain and frequency domain self-adaptive segmentation of signals, effectively solves the contradiction between high resolution of a frequency spectrum hole of the frequency spectrum and the quick and in-time detection requirement in frequency spectrum detection, and simultaneously, is used for estimating the energy distribution of time and frequency domains of main user signals. A simulation result verifies the effectiveness of the method.

Description

Self adaptation time-frequency hole detection method based on wavelet transformation

Technical field:

The present invention relates to a kind of time-frequency hole cognition technology that is applicable to the cognition wireless electrical domain; It adopts the method for handling based on wavelet transform signal; The time domain of wireless environment and the two-dimentional Energy distribution of frequency domain are carried out self-adapting detecting, with detection wireless environment time-frequency hole.

Background technology:

Wireless communication spectrum is a kind of limited precious resources, along with the information-intensive society rapid economy development, and the rare important bottleneck that becomes mobile communication and wireless access wide band technology realization day by day of frequency spectrum resource; What on the other hand, form sharp contrast with frequency spectrum resource shortage but is the extremely low of the existing availability of frequency spectrum.Under such background, cognitive radio technology inserts flexibly with it, can improve the availability of frequency spectrum greatly, effectively solves the deficient problem of frequency spectrum resource, is considered to one of potential key technology of next generation mobile communication system.

Cognitive radio system is not causing authorized user under the prerequisite of interference with in a kind of " chance mode " insertion authority frequency range, utilizes its interim idle frequency range dynamically.Realize this a pair of main user almost the noiseless of " transparent " take, require cognitive user reliably, the perception authorized user resumes and in time keep out of the way apace; On the other hand cognitive radio system require cognitive user reliably, cavity and taking again and again when detecting apace; Cognitive radio system will realize that high efficiency utilizes the purpose of frequency spectrum; Also should have the ability of estimating wireless frequency spectrum hole characteristic, and adjust transmission parameter (like data speed, transmission mode, transmission bandwidth etc.) adaptively based on frequency spectrum hole characteristic.Therefore, the frequency spectrum perception technology is the key technology of cognitive radio system.

Carried out number of research projects to cognitive radio frequency spectrum cognition technology and adaptive modulation technology both at home and abroad; Wherein roughly can be divided into following two big types to the frequency spectrum perception Study on Technology: (1) is collaborative to be detected: come to judge more accurately through the detection information of handling a plurality of perception users whether main user exists, these class methods will depend on complicated centralized control usually or distributed control realizes a plurality of perception users' detection information exchange; Whether (2) main user's transmitting terminal detects: detect cognitive user through analysis and receive in the signal and exist main subscriber signal to judge the user mode of frequency spectrum; These class methods can be divided into following four types again: 1) energy measuring method: this type algorithm need not known the prior information of authorization user signal usually; Can be readily applied to various Unknown Channel; Detect fast; Realize simply, but such algorithm dependence requires accurate estimating noise power, the noise power evaluated error can cause higher false alarm probability in the reality; 2) feature detection method: this type algorithm utilizes known main subscriber signal characteristic (like wave character, cyclostationarity etc.) that main CU frequency spectrum situation is detected; Can under strong interference environment, detect the frequency spectrum hole reliably; But this type algorithm requires to know in advance the characteristic information of main subscriber signal, and computation complexity is higher; 3) correlation detection algorithm: this type algorithm utilizes main subscriber signal sample value to have correlation this characteristics more much higher than noise sample value, realize the detection to authorized user, but this type algorithm is subject to the acute variation of signal correlation; 4) wavelet transformation frequency spectrum detection method: this type algorithm utilizes wavelet decomposition or WAVELET PACKET DECOMPOSITION can adjust the advantage of frequency spectrum detection resolution neatly; The realization multi-resolution spectrum detects; The sudden change feature extraction ability of perhaps utilizing small echo to change; Detect the sudden change of main CU-release frequency spectrum, but this type algorithm has the shortcoming of higher algorithm complex usually.

At present, existing both at home and abroad cognitive radio frequency spectrum cognitive method is primarily aimed at the detection in frequency domain " frequency spectrum hole ", and less from the angular detection " time-frequency hole " of time domain with the two-dimentional time frequency analysis of frequency domain, causes being difficult to estimate the time domain specification of idle frequency spectrum.This type frequency spectrum detecting method is in order to obtain higher frequency resolution; Need long period sampling at interval; On the other hand, in order in time to detect the frequency spectrum hole and main user resumes the frequency spectrum hole, require short time sampling interval; Therefore existing frequency spectrum detecting method also is subject to sampling interval long period and in time detects the contradiction between requiring fast except that having the said limitation of the joint of going up.

Summary of the invention:

The objective of the invention is to overcome the shortcoming of above-mentioned prior art; A kind of self adaptation time-frequency hole detection method based on wavelet transformation is provided; Only different from frequency domain detection frequency spectrum hole with existing cognitive radio frequency spectrum cognitive method, this programme adopts the wavelet transform signal Feature Extraction Technology, realizes the time domain and the frequency domain adaptive segmentation of signal; Solve sampling interval long period effectively and in time detect the contradiction between requiring fast, estimate the Energy distribution of the time-frequency domain of main subscriber signal simultaneously.

In addition, in the real network, the traffic carrying capacity owing to main user in the microcosmic time period has certain statistical law; Thereby frequency division Bu Tezheng also has certain statistical law at that time; Obtained the time-frequency distributions characteristic of main subscriber signal, can adjust sense cycle adaptively neatly better according to the statistical nature of each time-frequency piece; And the time-frequency hole predicted, insert opportunity loss and in time do not keep out of the way the harmful interference loss that causes main user thereby effectively reduce cognitive user.

Therefore, significant to cognitive radio system based on researching and analysing of time-frequency two-dimensional angle.

The objective of the invention is to solve through following technical scheme:

The time-frequency hole detection method that the present invention relates to is divided into three steps: the first step, and monitoring receives the segmentation of signal time domain to cognitive user earlier, guarantees in this segmentation, and each main CU frequency band continues constant; Second step, to each reception signal at times, adopt existing method, carry out Fourier transform and obtain each signal spectrum at times, this frequency spectrum is carried out wavelet transformation detecting each frequency range edge, thereby realize that frequency range divides.In the 3rd step, auto-adaptive time and frequency domain segmentation result in comprehensive preceding two steps are carried out analysis-by-synthesis to each time, whether judge the time free time.

Divide five parts that this method is introduced below; Elder generation of first introducing system model; Second portion and third part are introduced time domain and frequency domain segmentation method respectively; The time-frequency piece power spectral density of the 4th part after according to segmentation realizes that the time-frequency hole detects, and the 5th part has also proposed a kind of improving one's methods to above-mentioned time-frequency detection method.

During following method was introduced, the signal of employing was time-continuous signal, and this method can directly be generalized to the processing of discrete-time signal.

One. cognitive radio time-frequency sensory perceptual system model

To above 2 research contents, we adopt a wireless communication system with M potential cognitive user and N potential main user, suppose that the frequency range that this system takies is [f ₀, f _N], and bandwidth is B, the continuous signal that makes m potential cognitive user cognitive user receive is expressed as

$r_m\left(t\right)=\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\α}}_{n,m}{x}_n\left(t\right)+{\mathrm{\ω}}_m\left(t\right)---\left(1\right)$

Wherein, x _n(t) signal of representing n potential main user to send, and α _{N, m}Represent the signal gain between the individual potential cognitive user of n potential main user and m, ω _m(t) represent noise.Need to prove that each cognitive user adopts identical signal processing method in the introduction below, in order to simplify description, will no longer distinguish each cognitive user in the introduction of back, omits footnote m.For further easy analysis, we do following supposition to above-mentioned cognitive system, and analyze its supposition reasonability.

A1) because potential main CU corresponding band has temporal discontinuity, and the channel between main user and the cognitive user be quasistatic on this time period in each time period, promptly should interior α of time period _nBe constant;

A2) n potential main user of supposition taking frequency range in the duration, and the frequency band that takies can not change

A3) noise in the supposition reception signal is an additive white Gaussian noise.

In the wireless communication system, user's access base station process is through base station assigns up-downgoing frequency range and period; The communication of one whole can be divided into several short communication duration or time periods; Channel is approximate constant in the corresponding time period, and frequency band is also constant in the communication duration, and noise is assumed to additive white Gaussian noise usually in the signal and receive; Therefore above-mentioned hypothesis is reasonable, conforms to practical communication system.

Two. receive the segmentation of signal time-domain adaptive

Based on assumed condition a1); At each main CU frequency band in constant time period, can be similar to that to think that cognitive user receives average power signal constant, utilize to receive average power signal at times; Intend and adopt wavelet transformation detection signal average power catastrophe point; With each detection at edge at times of realization signal, thereby time-domain signal is divided into a plurality of time period continuous signals, concrete time-domain signal adaptive segmentation scheme is following:

1) calculates the reception average power signal

Ask its average power to received signal, formula specific as follows:

$r^{\′}\left(t\right)=\frac{1}{T}{\∫}_t^{t+T}{\left|r\left(t\right)\right|}^2\mathrm{dt}---\left(2\right)$

Time T can be chosen for 50 times of signal Transmission bit rate and got final product average period in the formula, so that can obtain average effect preferably.

2) multi-scale wavelet transformation

The wavelet function that makes wavelet transformation adopt is ψ (t), and the flexible function of this small echo can be expressed as

Know [22] by Wavelet Analysis Theory, cognitive user is received the wavelet transformation that signal carries out different scale, can regard wavelet function ψ as different scale _a(t),, can adopt the spectral density Ψ of elder generation to the two for simplifying computing with signal r ' convolution (t) _a(ω), R (ω) multiplies each other, and passes through Ψ again _a(ω) R (ω) inversefouriertransform obtains the wavelet transformation of different scale.

WT _ar′(t)＝F ^-1(Ψ _a(ω)R(ω)) (3)

3) wavelet transformation detects catastrophe point

Supposition by the front can know that the reception average power signal of cognitive user has similar step edge at the intersection of different periods, according to wavelet transformation character, at the edge moment of segmentation continuous signal { t _n, its wavelet transformation has local maximum, therefore can be through detecting the detection that cognitive user receives the wavelet transformation local maximum realization signal edge of average power signal.Be that edge moment sequencal estimation can obtain through following formula:

${\hat{t}}_n=\max \mathrm{im}{a}_t\left\{{\mathrm{WT}}_{a=2^j}{r}^{\&prime;}\left(t\right)\right\},t\&Element;({\mathrm{<figure-callout\; id="0"\; label="t"\; filenames="HDA0000129528020000012.png"\; state="\{\{state\}\}">t</figure-callout>}}_0,t_N)---\left(4\right)$

Wherein, (t ₀, t _N) expression cognitive user reception signal time corresponding section, adopt two minutes yardsticks, i.e. a=2 in this project ^j, j=0,1,2 ...

On the other hand; The then similar random pulses of the noise average power of time-domain signal (because the average period that average power signal adopts in calculating is shorter); By the characteristic of wavelet transformation maximum on multiple dimensioned; The wavelet transformation maximum at step edge does not increase with wavelet scale and changes, and the wavelet transformation local maximum of spike signal then increases with wavelet scale and reduces, therefore; The multi-scale wavelet transformation maximum of utilizing cognitive user to receive signal comes the time of main subscriber signal is carried out segmentation, has antinoise interference characteristic preferably.Detect anti-interference and noise effect ability in order further to improve the time slice edge, edge sequencal estimation constantly can be improved to

${\hat{t}}_n={\max \mathrm{ima}}_t\left\{\underset{j=0}{\overset{J}{\mathrm{\&Pi;}}}{\mathrm{WT}}_{a=2^j}{r}^{\&prime;}\left(t\right)\right\},t\&Element;({\mathrm{<figure-callout\; id="0"\; label="t"\; filenames="HDA0000129528020000012.png"\; state="\{\{state\}\}">t</figure-callout>}}_0,t_N)---\left(5\right)$

Consider to adopt the judgement thresholding that improves local extremum in order further to improve the reliability of signal time domain segmentation, to intend, but this thresholding determination methods can be brought certain hour segmentation error, needs to compromise consideration.

This thresholding b _THCan be taken as function Average, calculate as follows:

$b_{\mathrm{TH}}=\frac{1}{t_N-{\mathrm{<figure-callout\; id="0"\; label="t"\; filenames="HDA0000129528020000012.png"\; state="\{\{state\}\}">t</figure-callout>}}_0}{\&Integral;}_{t_0}^{t_N}\underset{j=0}{\overset{J}{\mathrm{\&Pi;}}}{\mathrm{WT}}_{a=2^j}{r}^{\&prime;}\left(t\right)\mathrm{dt}---\left(6\right)$

Three. the adaptive frequency domain segmentation

Based on the time domain segmentation of above-mentioned reception signal, can think that it is to continue and stablize constant that each main user takies the frequency spectrum situation in each time slice.Adopt the method in the existing document [13] to realize frequency domain adaptive segmentation in each time domain segmentation among the present invention, concrete grammar is following:

1) calculates the frequency domain power spectrum

Suppose to p at times signal carry out the frequency domain segmentation, wherein this at times time-domain signal can be expressed as:

$r^{\left(p\right)}\left(t\right)=\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_n{x}_n\left(t\right)+{\mathrm{\&omega;}}_m\left(t\right),t\&Element;[{\hat{t}}_p,{\hat{t}}_{p+1}]---\left(7\right)$

The first step is asked each power spectrum of time domain at times:

2) multi-scale wavelet transformation

Further this power spectrum is carried out the frequency-domain small wave conversion,

$W_{s=2^j}{S}_r^{\left(p\right)}\left(f\right)=S_r^{\left(p\right)}\left(f\right)*{\mathrm{\&phi;}}_{s=2^j}\left(f\right)---\left(9\right)$

Wherein

is basic flexible wavelet function; For the first derivative of a certain low pass smooth function, like Gaussian function.

3) the frequency domain catastrophe point detects

So the frequency spectrum segmentation edge sequence frequency of corresponding p time tick can be estimated to obtain through following formula:

${\hat{f}}_m^{\left(p\right)}={\max \mathrm{ima}}_f\left\{\underset{j=1}{\overset{J}{\mathrm{\&Pi;}}}\left[W_{s=2^j}{r}^{\&prime;}\left(t\right)\right]\right\},f\&Element;({\mathrm{<figure-callout\; id="0"\; label="f"\; filenames="HDA0000129528020000012.png"\; state="\{\{state\}\}">f</figure-callout>}}_0,f_N)---\left(10\right)$

Four. detect the time-frequency hole

1) estimating noise power spectrum

Main CU situation based on above-mentioned each time slice that has obtained, each sub-band can obtain whole time period (t ₀, t _N) and bandwidth [f ₀, f _N] in time take situation.Further, can be to m sub-frequency bands time in occupied p the time slice according to the computes power spectral density.

$P_n^{\left(p\right)}=\frac{1}{(f_{n+1}-f_n)}{\&Integral;}_{f_n}^{f_{n+1}}{S}_r^{\left(p\right)}\left(f\right)\mathrm{df}---\left(11\right)$

Because noise is an additive white Gaussian noise, therefore, noise is a constant in whole frequency domain power spectral density, can estimate to obtain through computes:

$N_{\mathrm{AWGN}}=\underset{n,p}{\min }{P}_n^{\left(p\right)}---\left(12\right)$

2) calculate each time-frequency block signal power spectrum

Average power signal spectrum density in p time slice on the m sub-frequency bands time can be modified to

$P_n^{\&prime;\left(p\right)}=P_n^{\left(p\right)}-N_{\mathrm{AWGN}}---\left(13\right)$

3) judge whether the time-frequency piece is idle

In theory, if some at times on the fast power spectral density of certain sub-frequency bands time-frequency greater than the white noise power spectral density, then this time is main CU, otherwise be the free time.Consider receive signal each at times the time interval limited, be not desirable constant still to this noise average power spectrum density that on average obtains at times, but have the stochastic variable of certain variance.Disturb for the fluctuation that suppresses the time average noise power spectrum, a higher power spectral density thresholding P is set as required _ThJudge whether each time is idle.

P _Th＝αN _AWGN (14)

Wherein factor-alpha＞1 can be provided with according to the computed reliability needs.

Five. time-frequency hole detection improvement

At system's acquisition frequency wider range; Perhaps under the situation of the cognitive user frequency range broad that need detect; Because the influence of multipath channel; Can cause cognitive user reception signal in the frequency range that system takies, to show as significantly frequency selectivity, and this frequency selectivity detect frequency domain separation formation interference for wavelet transformation; If system's acquisition frequency wider range on the other hand; The average power of noise is bigger; And some less sub-bands take or idle condition changes; Make that the change of average power of signal is less relatively, that is to say that interference of noise makes the method that adopts average power surge detection time domain waypoint produce erroneous judgement more easily or fails to judge.

To detect error in order overcoming this because system detects the time of the bigger generation of frequency range, to the invention allows for a kind of time-frequency hole detection improvement method, detailed process is as shown in Figure 1:

Utilize WAVELET PACKET DECOMPOSITION neatly signal decomposition to be become a plurality of signals that have than narrow bandwidth; Reception signal on each decomposition frequency band is carried out above-mentioned time-frequency adaptive segmentation respectively; At last each time domain segmentation of decomposing the signal on the frequency range is carried out comprehensively, can be obtained signal time-frequency hole detection more accurately.

Description of drawings:

Fig. 1 is the time-frequency hole detection improvement block diagram based on WAVELET PACKET DECOMPOSITION;

Fig. 2 (a) is the time-frequency hole detection system block diagram of handling based on wavelet transform signal;

Fig. 2 (b) is for receiving the signal filtering module frame chart;

Fig. 2 (c) is a time-domain adaptive segmentation module block diagram;

Fig. 2 (d) is a frequency domain adaptive segmentation module block diagram;

It during Fig. 2 (e) frequency hole comprehensive detection module frame chart;

Fig. 3 is the actual figure that takies of running time-frequency resource;

Fig. 4 is a signal time domain segmentation simulation result;

Fig. 5 is a signal frequency domain segmentation simulation result.

Embodiment:

Below in conjunction with accompanying drawing the present invention is done and to describe in further detail:

Realize being divided into following four steps based on the time-frequency hole self-adapting detecting method of wavelet transformation,

As shown in Figure 2:

The first step: the filtering of signal and sampling;

(1) time-domain signal is carried out continuous sampling;

(2) according to cognitive user sampling bandwidth (intercepting frequency bandwidth) size, whether decision adopts WAVELET PACKET DECOMPOSITION to the time-domain sampling signal.

Second step: through the time-domain adaptive segmentation of filtering signal

(1) sampling is asked the moving average power of signal by formula (2) to sampling time-domain signal (perhaps passing through the time-domain signal of WAVELET PACKET DECOMPOSITION);

(2) adopting the first derivative of a certain smooth function is wavelet function, and the moving average power signal in second step is carried out multi-scale wavelet transformation;

(3) ask its peak value limit that surpasses the threshold value that is provided by formula (6) by formula (5), this extreme point time corresponding point is time domain segmentation edge point.

The 3rd step: each is the signal frequency-domain adaptive segmentation at times

(1) asks each power spectral density of corresponding time-domain signal at times in the 4th step respectively;

(2) adopting the first derivative of a certain smooth function is wavelet function, respectively each power spectral density in the 5th step is carried out multi-scale wavelet transformation;

(3) ask the peak value limit of multi-scale wavelet transformation function by formula (10), this extreme point frequency points corresponding is frequency domain segmentation edge point;

The 4th step: estimate the spectral density on the time, and judge whether the free time;

(1) according to the time-frequency domain segmentation in the 4th step and the 7th step running time-frequency resource is divided into time, calculates the additive white Gaussian noise power spectral density by formula (12), and calculate the available signal power spectrum density of each time respectively by formula (13);

According to the given power spectral density thresholding of formula (14), judge taking or idle condition of each time.

1) experiment simulation

Emulation employing sub-carrier number is 1024 ofdm system, and subcarrier spacing is 1KHz, adopts the QPSK modulation; Suppose (corresponding 50 OFDM mark spaces at 50ms; Each OFDM mark space is 1ms) in running time-frequency resource take situation: in 0-10 OFDM interval, system takies whole 1024 number of sub-carrier, in 11-30 OFDM mark space; System takies 0-255 subcarrier and 513-767 subcarrier; In 31-50 OFDM mark space, system takies whole 0-255 subcarriers, and is as shown in Figure 3.The time-domain signal sample waveform of system is shown in Fig. 4 (a).

Intending the first derivative that adopts Gaussian function in the emulation is that the edge detects wavelet function, and this wavelet function is flexible to be that noise variance is taken as 0.1 on each subcarrier.

In this 50ms observation interval; Shown in following Fig. 4 of signal time-domain adaptive segmentation result (b); It is thus clear that, take the time point that situation changes based on the just corresponding actual frequency domain in the detected time domain of the average power signal of wavelet transformation edge, realized the time domain segmentation of signal well; First time domain edge as estimating is the 11st an OFDM symbol sampler starting point, the 31st the OFDM symbol sampler starting point in estimation point position, second time domain edge.

Fig. 5 has provided the power spectrum signal of corresponding second time domain segmentation and has adopted the power spectrum edge frequency estimated result of wavelet transformation; Wherein Fig. 5 (a) has provided corresponding signal power spectrum density, peak point among Fig. 5 (b) corresponding the power spectrum distortion point, i.e. frequency domain segmentation edge frequency: the 256th number of sub-carrier frequency; The 512nd number of sub-carrier frequency; The 768th number of sub-carrier frequency is divided into four frequency domain segmentations with 1024 subcarriers, takies situation with the frequency domain of segmentation in second o'clock and conforms to fully.

The above only is preferred embodiment of the present invention, is not the present invention is done any pro forma restriction; Though the present invention discloses as above with preferred embodiment; Yet be not in order to limiting the present invention, anyly be familiar with the professional and technical personnel, in not breaking away from technical scheme scope of the present invention; When the method for above-mentioned announcement capable of using and technology contents are made a little change or be modified to the equivalent embodiment of equivalent variations; In every case be the content that does not break away from technical scheme of the present invention, to any simple modification, equivalent variations and modification that above embodiment did, still belong in the scope of technical scheme of the present invention according to technical spirit of the present invention.

Claims (7)

1. based on the self adaptation time-frequency hole detection method of wavelet transformation; It is characterized in that; Adopt wavelet transformation technique adaptively time-continuous signal or discrete-time signal to be carried out the time-frequency domain segmentation, and the free time of judging each time whether, comprises the steps:

(1) monitoring receives the segmentation of signal time domain to cognitive user earlier, guarantees that in this segmentation each main CU frequency band continues constant;

(2) to each reception signal at times, adopt existing method, carry out Fourier transform and obtain each signal spectrum at times, this frequency spectrum is carried out wavelet transformation detecting each frequency range edge, thereby realize that frequency range divides;

(3) auto-adaptive time and the frequency domain segmentation result in comprehensive preceding two steps carried out analysis-by-synthesis to each time, whether judges the time free time.

2. self adaptation time-frequency as claimed in claim 1 hole detection method; It is characterized in that: this method adopts to the average power that receives signal carries out wavelet transformation; The adaptive time domain segmentation that achieves a butt joint and collect mail number guarantees that in each time slice, the spectrum structure of signal is constant.

3. self adaptation time-frequency as claimed in claim 1 hole detection method is characterized in that: this method adopts WAVELET PACKET DECOMPOSITION or other filtering receiving handling method to carry out frequency-division section filtering to received signal and receives processing.

4. self adaptation time-frequency as claimed in claim 1 hole detection method is characterized in that:

Cognitive radio time-frequency sensory perceptual system model adopts a wireless communication system with M potential cognitive user and N potential main user, supposes that the frequency range that this system takies is [f ₀, f _N], and bandwidth is B, the continuous signal that makes m potential cognitive user cognitive user receive is expressed as

$r_m\left(t\right)=\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_{n,m}{x}_n\left(t\right)+{\mathrm{\&omega;}}_m\left(t\right)---\left(1\right)$

Wherein, x _n(t) signal of representing n potential main user to send, and α _{N, m}Represent the signal gain between the individual potential cognitive user of n potential main user and m, ω _m(t) represent noise; Each cognitive user adopts identical signal processing method in the introduction below, in order to simplify description, will no longer distinguish each cognitive user in the introduction of back, omits footnote m; We do following supposition to above-mentioned cognitive system, and analyze its supposition reasonability;

A1) because potential main CU corresponding band has temporal discontinuity, and the channel between main user and the cognitive user be quasistatic on this time period in each time period, promptly should interior α of time period _nBe constant;

A2) n potential main user of supposition taking frequency range in the duration, and the frequency band that takies can not change;

A3) noise in the supposition reception signal is an additive white Gaussian noise;

In the wireless communication system, user's access base station process is through base station assigns up-downgoing frequency range and period; The communication of one whole can be divided into several short communication duration or time periods; Channel is approximate constant in the corresponding time period, and frequency band is also constant in the communication duration, and noise is assumed to additive white Gaussian noise usually in the signal and receive; Therefore above-mentioned hypothesis is reasonable, conforms to practical communication system;

Receive the segmentation of signal time-domain adaptive: based on assumed condition a1); At each main CU frequency band in constant time period; Can be similar to and think that cognitive user reception average power signal is constant, utilize to receive average power signal at times, intend and adopt wavelet transformation detection signal average power catastrophe point; With each detection at edge at times of realization signal, thereby time-domain signal is divided into a plurality of time period continuous signals;

Adaptive frequency domain segmentation:, can think that it is to continue and stablize constant that each main user takies the frequency spectrum situation in each time slice based on the time domain segmentation of above-mentioned reception signal;

Detect the time-frequency hole

1) estimating noise power spectrum

Main CU situation based on above-mentioned each time slice that has obtained, each sub-band can obtain whole time period (t ₀, t _N) and bandwidth [f ₀, f _N] in time take situation; Further, to m sub-frequency bands time in occupied p the time slice; According to the computes power spectral density;

$P_n^{\left(p\right)}=\frac{1}{(f_{n+1}-f_n)}{\&Integral;}_{f_n}^{f_{n+1}}{S}_r^{\left(p\right)}\left(f\right)\mathrm{df}---\left(11\right)$

Because noise is an additive white Gaussian noise, therefore, noise is a constant in whole frequency domain power spectral density, estimates to obtain through computes:

$N_{\mathrm{AWGN}}=\underset{n,p}{\min }{P}_n^{\left(p\right)}---\left(12\right)$

2) calculate each time-frequency block signal power spectrum

Average power signal spectrum density in p time slice on the m sub-frequency bands time can be modified to

$P_n^{\&prime;\left(p\right)}=P_n^{\left(p\right)}-N_{\mathrm{AWGN}}---\left(13\right)$

3) judge whether the time-frequency piece is idle

In theory, if some at times on the fast power spectral density of certain sub-frequency bands time-frequency greater than the white noise power spectral density, then this time is main CU, otherwise be the free time; Consider receive signal each at times the time interval limited, be not desirable constant still to this noise average power spectrum density that on average obtains at times, but have the stochastic variable of certain variance; Disturb for the fluctuation that suppresses the time average noise power spectrum, a higher power spectral density thresholding P is set as required _ThJudge whether each time is idle;

P _Th＝α _NAWGN (14)

Wherein factor-alpha＞1 can be provided with according to the computed reliability needs.

Time-frequency hole detection improvement

At system's acquisition frequency wider range; Perhaps under the situation of the cognitive user frequency range broad that need detect; Because the influence of multipath channel; Can cause cognitive user reception signal in the frequency range that system takies, to show as significantly frequency selectivity, and this frequency selectivity detect frequency domain separation formation interference for wavelet transformation; If system's acquisition frequency wider range on the other hand; The average power of noise is bigger; And some less sub-bands take or idle condition changes; Make that the change of average power of signal is less relatively, that is to say that interference of noise makes the method that adopts average power surge detection time domain waypoint produce erroneous judgement more easily or fails to judge.

5. self adaptation time-frequency as claimed in claim 4 hole detection method, it is characterized in that: said time-domain signal adaptive segmentation is following:

1) calculates the reception average power signal

Ask its average power to received signal, formula specific as follows:

$r^{\&prime;}\left(t\right)=\frac{1}{T}{\&Integral;}_t^{t+T}{\left|r\left(t\right)\right|}^2\mathrm{dt}---\left(2\right)$

Average period, time T can be chosen for 50 times of the signal Transmission bit rate in the formula, so that can obtain average effect preferably;

2) multi-scale wavelet transformation

The wavelet function that makes wavelet transformation adopt is ψ (t), and the flexible function of this small echo can be expressed as

Know by Wavelet Analysis Theory, cognitive user is received the wavelet transformation that signal carries out different scale, can regard wavelet function ψ as different scale _a(t),, can adopt the spectral density Ψ of elder generation to the two for simplifying computing with signal r ' convolution (t) _a(ω), R (ω) multiplies each other, and passes through Ψ again _a(ω) R (ω) inversefouriertransform obtains the wavelet transformation of different scale;

WT _ar′(t)＝F ^-1(Ψ _a(ω)R(ω)) (3)

3) wavelet transformation detects catastrophe point

Supposition by the front can know that the reception average power signal of cognitive user has similar step edge at the intersection of different periods, according to wavelet transformation character, at the edge moment of segmentation continuous signal { t _n, its wavelet transformation has local maximum, so can be through detecting the detection that cognitive user receives the wavelet transformation local maximum realization signal edge of average power signal, and promptly edge moment sequencal estimation can obtain through following formula:

${\hat{t}}_n=\max \mathrm{im}{a}_t\left\{{\mathrm{WT}}_{a=2^j}{r}^{\&prime;}\left(t\right)\right\},t\&Element;(t_0,t_N)---\left(4\right)$

Wherein, (t ₀, t _N) expression cognitive user reception signal time corresponding section, adopt two minutes yardsticks, i.e. a=2 in this project ^j, j=0,1,2 ...

6. self adaptation time-frequency as claimed in claim 4 hole detection method is characterized in that: the frequency domain adaptive segmentation is according to following steps in each time domain segmentation of said realization:

1) calculates the frequency domain power spectrum

Suppose to p at times signal carry out the frequency domain segmentation, wherein this at times time-domain signal can be expressed as:

$r^{\left(p\right)}\left(t\right)=\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_n{x}_n\left(t\right)+{\mathrm{\&omega;}}_m\left(t\right),t\&Element;[{\hat{t}}_p,{\hat{t}}_{p+1}]---\left(7\right)$

The first step is asked each power spectrum of time domain at times:

2) multi-scale wavelet transformation

Further this power spectrum is carried out the frequency-domain small wave conversion,

$W_{s=2^j}{S}_r^{\left(p\right)}\left(f\right)=S_r^{\left(p\right)}\left(f\right)*{\mathrm{\&phi;}}_{s=2^j}\left(f\right)---\left(9\right)$

Wherein

is basic flexible wavelet function; For the first derivative of a certain low pass smooth function, like Gaussian function;

3) the frequency domain catastrophe point detects

So the frequency spectrum segmentation edge sequence frequency of corresponding p time tick can be estimated to obtain through following formula:

${\hat{f}}_m^{\left(p\right)}={\max \mathrm{ima}}_f\left\{\underset{j=1}{\overset{J}{\mathrm{\&Pi;}}}\left[W_{s=2^j}{r}^{\&prime;}\left(t\right)\right]\right\},f\&Element;(f_0,f_N)---\left(10\right).$

7. self adaptation time-frequency as claimed in claim 1 hole detection method; It is characterized in that; Time-frequency analysis-by-synthesis estimating noise power spectral method: according to time domain and frequency domain segmentation result; Carry out each time internal power spectrum and carry out analysis-by-synthesis, more accurately estimate to obtain noise power spectral density, thereby for to judge whether the free time provides decision threshold accurately to time.

Images