CN104267413B - Lifting Wavelet dual threshold Denoising Algorithm based on signal intensity self adaptation TABU search - Google Patents

Abstract

The present invention relates to a kind of Lifting Wavelet dual threshold Denoising Algorithm based on signal intensity self adaptation TABU search, its technical characteristics is: utilize differential coherence Cumulate algorithm to capture faint received global navigation satellite system signal；Faint received global navigation satellite system signal is analyzed by application Lifting Wavelet decomposition；Dual threshold in Lifting Wavelet denoising is selected to be optimized by application signal intensity self adaptation tabu search algorithm；Utilize optimized dual threshold that the faint received global navigation satellite system signal of capture is carried out post processing.The present invention is reasonable in design, Weak GNSS signal is captured by differential coherence Cumulate algorithm, the Lifting Wavelet dual threshold denoising method of application signal intensity self adaptation TABU search optimization carries out post processing to the Weak GNSS signal obtained, achieve relatively low calculating time cost and higher accuracy, improve the signal to noise ratio of signal output.

Description

Lifting Wavelet dual threshold Denoising Algorithm based on signal intensity self adaptation TABU search

Technical field

The invention belongs to Weak Signal Processing technical field, especially one search based on the taboo of signal intensity self adaptation The Lifting Wavelet dual threshold Denoising Algorithm of rope.

Background technology

In recent years, in order to obtain more accurate navigation information, such as height, speed and position, worldwide navigation is defended Star system (GNSS) technology is widely used in military field, and vehicle mounted guidance, personal hand-held terminal and Multiple civil area all plays very important role.But, at ratio under relatively rugged environment, such as township Village, forest, valley and indoor environment, owing to the reason of propagation loss and multipath fading makes the letter that receives Number signal to noise ratio the lowest, compared with open environment, signal can produce the decay of 15dB-20dB.It addition, The power of aeronautical satellite small-signal is far below the scope of application of general operation of receiver power.Therefore, tradition GNSS signal receiver can not capture gps signal in the environment of high request, not to mention follow the tracks of and fixed Position.In sum, need to increasingly focus on the capture of Weak GNSS signal, and receiver is at capture amplitude Small-signal aspect that is relatively low and that be embedded in noise requires the sensitiveest, meanwhile, accumulated time, higher The lifting of signal to noise ratio and sensitivity is all to need research and the direction paid close attention to.

The method of Weak absorption is different from the detection of general signal, and traditional signal detecting method is typically paid close attention to Reduction and the raising of signal to noise ratio in noise.But when processing small-signal, these conventional methods often lost efficacy, Because conventional method can not be effectively improved the signal to noise ratio of the small-signal corroded by noise, and at traditional letter Number amplify during, small-signal easily torsional deformation.It addition, what some traditional small-signals obtained Method (such as low frequency phase sensitive filter) needs some to have the specific information of OFF signal, such as phase place, Frequency etc..Accordingly, it would be desirable to invent some new methods obtaining small-signal and use it for solving practical problems.

Summary of the invention

It is an object of the invention to overcome the deficiencies in the prior art, it is provided that one is reasonable in design, precision is high and the time The Lifting Wavelet dual threshold Denoising Algorithm based on signal intensity self adaptation TABU search of low cost.

The present invention solves existing technical problem is that and takes techniques below scheme to realize:

A kind of Lifting Wavelet dual threshold Denoising Algorithm based on signal intensity self adaptation TABU search, including following Step:

Step 1, utilize differential coherence Cumulate algorithm capture faint received global navigation satellite system signal；

Faint received global navigation satellite system signal is analyzed by step 2, application Lifting Wavelet decomposition；

Dual threshold in Lifting Wavelet denoising is selected by step 3, application signal intensity self adaptation tabu search algorithm It is optimized；

Step 4, utilize optimized dual threshold to capture faint received global navigation satellite system signal carry out after locate Reason.

And, the implementation method of described step 1 is: obtain small-signal by differential coherence accumulation, then Accumulation results is inputted Wavelet Denoising Method wave filter to find its code phase and Doppler frequency shift, thus complete capture Process.

And, the concrete step that processes of described step 2 is:

(1) decomposition step, by original signal sequence S_jIt is decomposed into two subdivisions according to parity even_j-1And odd_j-1:

(odd_j-1,even_j-1)=Split (S_j)

(2) prediction steps, is predicted by following prediction machine:

d_j-1=odd_j-1-P(even_j-1)

Wherein, P (even_j-1) it is to predict odd number value, specifying information d by even item_j-1Pass through odd_j-1With in advance Survey P (even_j-1Between), error represents；

(3) update step, update average as the following formula:

S_j-1=even_j-1+ U (d_j-1)。

And, the concrete step that processes of described step 3 is:

(1) upper threshold and lower limit T ' are supposed_minWith T '_max

(2) the initial interval of threshold optimization is set to [T_min,T_max], by upper limit T_maxIt is set to T_max=3 σ, Wherein σ is estimated as: σ ≈ Mid/0.6745, in formula, Mid be Decomposition order minimum time wavelet coefficient by Median after numerical ordering, lower limit T_minIt it is the absolute value of minimum coefficient than 0 in jth layer wavelet decomposition；

(3) at T '_minWith T '_maxBetween produce a candidate solution subset, then according to following fitness function It is ranked up:

$S=\mathrm{\α}\×\lg \[\frac{\underset{n}{\Σ}{f}_d^2\left(n\right)}{\underset{n}{\Σ}{\[f_d\left(n\right)-f\left(n\right)\]}^2}\]$

Wherein, α is the auto-adaptive parameter that the strength information according to faint received global navigation satellite system signal is chosen, f_dFor Reconstruction signal after denoising, f is reference signal；

$\mathrm{\α}=\sqrt{{\mathrm{\σ}}_{\mathrm{\ω}}^2\left(t\right)+{\mathrm{\σ}}_n^2}$

Wherein, σ_ωT () is the root-mean-square value of signal,For system average noise power；

(4) scan for by signal intensity self adaptation TS algorithm, be modified current solution obtaining feasible Solving, if moving to its neighbouring solution to be not better than the most current optimal solution, the most adjacent solution will be accepted；Then draw Enter to avoid list and avoid circulation；

(5) step (3) is returned until meeting stopping criterion；

(6) utilize equation below calculate optimal threshold:

T′_max=y_1j×10⁰+y_2j×10^-1+y_3j×10^-2+y_4j×10^-3

T′_{m in}=y_5j×10⁰+y_6j×10^-1+y_7j×10^-2+y_8j×10^-3。

And, the concrete processing method of described step 4 is: at optimal double threshold value [T '_min,T′_max] be determined after, Kth wavelet coefficient W in j layer wavelet decomposition_j,kProcess as the following formula:

$W_{j,k}^{\&prime;}=\left\{\begin{array}{c}0,\left|W_{j,k}\right|<T_{\min }^{\&prime;};\\ \frac{T_{\max }^{\&prime;}}{T_{\max }^{\&prime;}-T_{\min }^{\&prime;}}\left(\left|W_{j,k}\right|-T_{\min }^{\&prime;}\right)\mathrm{sgn}\left(W_{j,k}\right),T_{\min }^{\&prime;}\&le;\left|W_{j,k}\right|\&le;T_{\max }^{\&prime;}\\ W_{j,k},\left|W_{j,k}\right|>T_{\max }^{\&prime;}.\end{array}\right..$

Advantages of the present invention and good effect be:

1, the present invention is during capture small-signal, uses differential coherence accumulation to obtain small-signal, energy Enough reducing Squared Error Loss, reduce the amplification of noise, the raising to signal to noise ratio is bigger simultaneously.

2, the present invention is by small-signal capture and the combination of Wavelet Denoising Method, signal to noise ratio can be brought up to some Specific degree, utilizes Wavelet noise-eliminating method to improve output signal-to-noise ratio.

3, the present invention uses tabu search algorithm, it is possible to the optimization feature of simulating human memory function, and passes through Local neighbor seaching mechanism and corresponding taboo criterion avoid roundabout search, it is possible to discharge some outstanding taboos Avoid the multiformity that state ensures to have efficient search, can suppress to fall into lacking of easy precocious in genetic algorithm Fall into, reach final optimization.

4, the present invention is reasonable in design, captures Weak GNSS signal, application letter by differential coherence Cumulate algorithm The Lifting Wavelet dual threshold denoising method that number Self-adaptive strength TABU search the optimizes Weak GNSS signal to obtaining Carry out post processing, it is achieved that relatively low calculates time cost and higher accuracy, improve signal output Signal to noise ratio.

Accompanying drawing explanation

Fig. 1 is the signal intensity self adaptation TABU search process schematic of the present invention；

Fig. 2 is relation comparison diagram between the different wavelet basis wavelet decomposition number of plies and output signal-to-noise ratio；

Fig. 3 is present invention denoising effect comparison diagram compared with primary signal；

Fig. 4 is the output signal-to-noise ratio comparison diagram under the present invention and other three kinds of algorithms before and after noise reduction.

Detailed description of the invention

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

A kind of Lifting Wavelet dual threshold Denoising Algorithm based on signal intensity self adaptation TABU search, including following Step:

Step 1, utilize differential coherence Cumulate algorithm capture faint received global navigation satellite system signal.

In the presence of a harsh environment, such as genuine and valley, prior art proposed different technological means and overcame A difficult problem for faint GLONASS (GNSS) signal capture.In order to improve the sensitivity of receiver, And obtaining Weak GNSS signal more accurately, traditional algorithm includes coherent accumulation (COH), non-coherent accumulation (NCOH) and differential coherence accumulation (DFC).In terms of Weak GNSS signal capture, COH is a kind of straight Connect and effective method, the signal to noise ratio of the signal detected can be improved.But, the navigation data of 50bps and Frequency shift (FS) will double to limit the time of coherent accumulation.Use the performance of differential coherence accumulation (DFC) algorithm The ratio performance mean height 1.5dB of non-coherent accumulation (NCOH), DFC algorithm can obtain the processing gain of maximum.

Therefore, this step utilizes differential coherence accumulation (DFC) to capture faint received global navigation satellite system signal. During capture small-signal, first implement differential coherence accumulation and obtain small-signal, then will accumulation Result input Wavelet Denoising Method wave filter, to find its code phase and Doppler frequency shift, has captured subsequently.Differential Coherent accumulation can reduce Squared Error Loss, and it is conjugated multiplication in adjacent coherent integrator catch cropping.Therefore, this Method is less to the amplification of noise, and the raising to signal to noise ratio is bigger.

The digital intermediate frequency signal of global navigation satellite system GNSS can be represented as:

$\begin{array}{c}{S}_{IF}\left(t\right)=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}2P_i{D}_i\left(t-{\mathrm{\&tau;}}_i\right)C_i\left(t-{\mathrm{\&tau;}}_i\right)\&CenterDot;\exp j\left(2\mathrm{\&pi;}\left(f_{IF}+f_{di}\right)t+{\mathrm{\&phi;}}_i\right)\\ +\mathrm{\&xi;}\left(t\right)\end{array}$

Wherein, P_iIt is signal amplitude, D_iIt is navigation data, C_iIt is the C/A when moment k with time delay τ Spreading code.Having the dependency between the satellite-signal of same pseudo-random noise PRN and local C/A code can To be expressed as:

$\left\{\begin{array}{c}{s}_I\left(t\right)=S_{IF}\left(t\right)\hat{C}(t-\widehat{\mathrm{\&tau;}})cos({\mathrm{\&omega;}}_{IF}t+{\widehat{\mathrm{\&omega;}}}_{d}t)\\ s_Q\left(t\right)=S_{IF}\left(t\right)\hat{C}(t-\widehat{\mathrm{\&tau;}})sin({\mathrm{\&omega;}}_{IF}t+{\widehat{\mathrm{\&omega;}}}_{d}t)\end{array}\right.$

Wherein, s_IT () is in-phase component part, and s_QT () is quadrature component part.

If correlation time is N number of cycle, then correlated results is:

$S_I=\frac{1}{2}{N}_{S}ADR\left(\mathrm{\&Delta;}\mathrm{\&tau;}\right)\sin c\left({\mathrm{\&Delta;\&omega;}}_D{T}_C\right)cos({\mathrm{\&Delta;\&omega;}}_D{T}_C+\mathrm{\&phi;})+n_{2,I}$

$S_Q=\frac{1}{2}{N}_{S}ADR\left(\mathrm{\&Delta;}\mathrm{\&tau;}\right)\sin c\left({\mathrm{\&Delta;\&omega;}}_D{T}_C\right)sin({\mathrm{\&Delta;\&omega;}}_D{T}_C+\mathrm{\&phi;})+n_{2Q}$

Wherein, A represents that signal amplitude, R (τ) are auto-correlation function, noise n_2,IWith n_2,QGaussian distributedN () represents Gauss distribution.

When using the accumulation of real differential, the differential cumulative process of a signal is as follows:

$\begin{array}{c}{\mathrm{DF}}_k=S_{k,l}{S}_{k-a,I}+S_{k,Q}{S}_{k-a,Q}\&ap;\frac{1}{4}{N}_S^2{A}^2{R}^2\left(\mathrm{\&Delta;}\mathrm{\&tau;}\right){\mathrm{sinc}}^2\left({\mathrm{\&Delta;\&omega;}}_d{N}_S{T}_C\right)\\ +n_{2,I,k}\&times;n_{2,I,k-1}+n_{2,Q,k}\&times;n_{2,Q,k-1}\end{array}$

If differential cumulative number is M step, then result is:

$DF=\underset{k=1}{\overset{M}{\mathrm{\&Sigma;}}}{\mathrm{DF}}_k=\frac{1}{4}{\mathrm{MN}}_S^2{A}^2{R}^2\left(\mathrm{\&Delta;}\mathrm{\&tau;}\right){\mathrm{sinc}}^2\left({\mathrm{\&Delta;\&omega;}}_d{N}_S{T}_C\right)+n_{3,I}+n_{3,Q}$

Wherein Δ τ_kWith Δ τ_k-1Represent Delay Estima-tion error during moment k and k-1, Δ ω respectively_d,kWith Δ ω_d,k-1Point Do not represent the Doppler frequency estimation error of moment k and k-1, Δ ω_dRepresent average to moment k from moment k-1 Doppler frequency estimation error.May certify that and have when code phase and Doppler frequency shift are perfectly aligned Δτ_k=Δ τ_k-1=0 and Δ ω_d,k=Δ ω_d,k-1=0.

$max\left\{DF\right\}\&ap;\frac{1}{4}{\mathrm{MN}}_S^2{A}^2$

Wherein, noise n_3,IAnd n_3,QGaussian distributedN () represents Gauss distribution, and γ is big In the constant coefficient of zero, its value depends on the time of integration.

Faint received global navigation satellite system signal is analyzed by step 2, application Lifting Wavelet decomposition.

This step application Lifting Wavelet is decomposed Weak GNSS signal is analyzed including following three step:

(1) decomposition step:

In decomposition, original signal sequence S_jIt is broken down into two subdivisions even according to parity_j-1With odd_j-1

(odd_j-1,even_j-1)=Split (S_j)

The lazy wavelet transformation that this step is referred to as in lifting scheme.Signal decomposition is simply two by this step the most simply Part, can not promotion signal.Next step promoted will identify that the two sequence is to reduce its dependency.

(2) prediction steps:

Prediction machine is defined as

d_j-1=odd_j-1-P(even_j-1)

Wherein, P (even_j-1) it is to predict odd number value by even item, and specifying information d_j-1Pass through odd_j-1With Prediction P (even_j-1Between), error represents.The dual lifting that this step is referred to as in lifting scheme.Phase when signal When closing property is the highest, it was predicted that be very effective.

(3) step is updated:

In order to make similarity signal S_j-1Maintain primary signal S_jSome characteristics, update average the most as the following formula:

S_j-1=even_j-1+ U (d_j-1)

The original lifting that this step is referred to as in lifting scheme.

If three steps of above-mentioned decomposition are all used for decomposed similarity signal S_j-1, then in a number of iteration time Can obtain primary signal S after number_jMultilamellar decompose.It is found that use lifting scheme to carry out small echo One of the sharpest edges decomposed are that wavelet transformation can be decomposed into some basic steps and carries out, and theirs is inverse Conversion is easier to find.

For the setting identical with wavelet filter, lifting process can improve the speed of wavelet transformation doublely. The computation complexity of wavelet transformation is defined as when ground floor decomposes the multiplication calculated required for a pair coefficient Number and addition number of times.For lifting scheme, it is assumed that the length of predicted operation coefficient and the length of renewal coefficient of performance Degree M and N respectively.Can obtain:

$d\left(n\right)=odd\left(n\right)-\underset{k=1}{\overset{M}{\mathrm{\&Sigma;}}}p\left(k\right)even(n-M_d+k-1)$

$S\left(n\right)=even\left(n\right)-\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}u\left(k\right)d(n-N_d+k-2)$

Wherein, M_d=M/2-1, N_d=N/2-1.

Can draw from above formula, the computation complexity of lifting scheme is 2 (M+N), including (M+N) secondary taking advantage of The addition that method is secondary with (M+N).It addition, the computation complexity of wavelet transform (DWT) is 4 (M+N) +2。

Therefore, for longer wave filter, the amount of calculation of application lifting scheme is about the one of traditional algorithm Half.

Dual threshold in Lifting Wavelet denoising is selected by step 3, application signal intensity self adaptation tabu search algorithm It is optimized.

The present invention is small-signal capture and Wavelet Denoising Method to be combined, and wherein Wavelet Denoising Method extensively should For a lot of fields.Signal to noise ratio can be brought up to some specific degree by Wavelet Denoising Method, therefore at satellite Navigation application also becomes a study hotspot.When Wavelet Denoising Method proposes first for target be reduce White noise, is then applied to ultrasonic signal and obtains and noise suppressed aspect.As for field of satellite navigation, carry Go out a kind of method based on wavelet transform (DWT) and be applied to the capture of GNSS signal.Further Ground, utilizes Wavelet noise-eliminating method to improve output signal-to-noise ratio and has been also applicable in weak GPS signals acquisition detection Post processing aspect.

Tabu search algorithm is a kind of overall situation Neighborhood-region-search algorithm, and the method simulates the optimization of human memory's function Feature, and avoid roundabout search by local neighbor seaching mechanism and corresponding taboo criterion.Meanwhile, logical Crossing the standard of breaking a taboo, it can also discharge the multiformity that some outstanding taboo states ensure to have efficient search. It addition, TABU search can suppress to fall into the defect of easy precocious in genetic algorithm, reach final Optimization.

Wavelet Denoising Method mode includes following two mode: hard-threshold mode and soft-threshold mode.The former is by low All coefficients in bottom threshold are set to zero, and other coefficient all keeps constant；The latter is to will be less than threshold value equally All coefficients of lower limit are set to zero, but other coefficient shrinks to zero.Scheme, heuristic calculation are determined for threshold value Method, after the iteration carrying out certain step number, can choose the threshold value of optimum in theory.

Weak GNSS signal comprises high-frequency components and low frequency component.Owing to the ratio of both compositions is not solid Fixed, if therefore easy choice hard-threshold mode or soft-threshold mode are certain by inevitably losing The information of frequency range, because they delete high-frequency components the most rambunctiously.Therefore, two kinds above-mentioned Algorithm all can not obtain gratifying performance during Wavelet Denoising Method.

The present invention apply signal intensity self adaptation Tabu search algorithm carry out Wavelet Denoising Method.Due to tabu search algorithm Being a kind of heuritic approach, it is prior to the heuritic approach combined by random algorithm and local investigation algorithm. The signal intensity self adaptation Tabu search algorithm process of the present invention is as follows:

(1) upper threshold and lower limit T ' are supposed_minWith T '_maxIt is defined as four position effective digitals, wherein three positions After arithmetic point, such as [y₅,y₆,y₇,y₈], [y₁,y₂,y₃,y₄]。

(2) the initial interval of threshold optimization is set to [T_min,T_max], upper limit T_maxIt is set to:

T_max=3 σ

In formula, σ is estimated as σ ≈ Mid/0.6745.

Mid be Decomposition order minimum time wavelet coefficient by the median after numerical ordering.

Lower limit T_minIt it is the absolute value of minimum coefficient than 0 in jth layer wavelet decomposition.

To retrain T ' as follows_minWith T '_maxInitialize:

T_min≤T′_min＜ T '_max≤T_max

(3) at T '_minWith T '_maxBetween produce a candidate solution subset, then according to following fitness function It is ranked up:

$s=\mathrm{\&alpha;}\&times;\lg \&lsqb;\frac{\underset{n}{\&Sigma;}{f}_d^2\left(n\right)}{\underset{n}{\&Sigma;}{\&lsqb;f_d\left(n\right)-f\left(n\right)\&rsqb;}^2}\&rsqb;$

Wherein, α is the auto-adaptive parameter that the strength information according to faint received global navigation satellite system signal is chosen, It can be according to the value of the actual information self-adaptative adjustment fitness function of the small-signal of capture so that adapt to Degree function can be made for different small-signals and judging more accurately；f_dFor the reconstruction signal after denoising, f For reference signal, and the value of s is the biggest, represents that fitness is the highest；

$\mathrm{\&alpha;}=\sqrt{{\mathrm{\&sigma;}}_{\mathrm{\&omega;}}^2\left(t\right)+{\mathrm{\&sigma;}}_n^2}$

Wherein, σ_ωT () is the root-mean-square value of signal,For system average noise power.

(4) according to the process of the TABU search of signal intensity self adaptation shown in Fig. 1, current solution is carried out once letter Single correction obtains feasible solution, and such process is referred to as a moved further.If moving to its neighbouring solution wait Choosing solves T* and is not better than current optimal solution, and in order to avoid local optimum, candidate solution T* will be accepted regardless of it Whether it is optimal solution.Then introduce taboo list and avoid circulation.Taboo list stores all can not be by It is accepted as the mobile step currently solved.The mobile step meeting taboo rule will be stored in taboo list.Taboo The use of list reduces the probability of circulation, because it prevent that return in the recent period in certain iterative steps The solution just accessed.

(5) (3) are returned until meeting stopping criterion.

(6) utilize equation below calculate optimal threshold:

T′_max=y_1j×10⁰+y_2j×10^-1+y_3j×10^-2+y_4j×10^-3

T′_min=y_5j×10⁰+y_6j×10^-1+y_7j×10^-2+y_8j×10^-3。

Step 4, utilize optimized dual threshold to capture faint received global navigation satellite system signal carry out after locate Reason.

Optimal double threshold value [T ' in step 3_min,T′_max] be determined after, as the following formula by kth in jth layer wavelet decomposition Individual wavelet coefficient W_j,k:

$W_{j,k}^{\&prime;}=\left\{\begin{array}{c}0,\left|W_{j,k}\right|<T_{\min }^{\&prime;};\\ \frac{T_{\max }^{\&prime;}}{T_{\max }^{\&prime;}-T_{\min }^{\&prime;}}\left(\right|W_{j,k}|-T_{\min }^{\&prime;})\mathrm{sgn}\left(W_{j,k}\right),T_{\min }^{\&prime;}\&le;\left|W_{j,k}\right|\&le;T_{\max }^{\&prime;}\\ W_{j,k},\left|W_{j,k}\right|>T_{\max }^{\&prime;}.\end{array}\right.;$

Achieved by above four steps be applied to Weak GNSS signal capture based on signal intensity self adaptation The Lifting Wavelet dual threshold Denoising Algorithm that TABU search optimizes, hence it is evident that reduce time complexity and improve Make an uproar accuracy.

In order to illustrate effect of the present invention, the mode using Computer Simulation below is faint to being applied to The present invention of GNSS signal capture is modeled, and achieves the simulation to real scene by assignment.Numeral Intermediate-freuqncy signal is as the input signal of emulation, and detailed process point following four aspect is analyzed:

(1) wavelet basis and the selection of Decomposition order

As shown in Figure 2, when decomposition level rises to 3, signal to noise ratio keeps stable and even performs better than. Haar wavelet transform base is put up the best performance in all candidate's wavelet basiss, and owing to signal is square wave, has minor fluctuations Wavelet basis more applicable, therefore select Haar wavelet transform base as in inventive algorithm application wavelet basis.

It addition, also contains the discussion selected for Decomposition order in fig. 2.If Decomposition order is too much, Owing to calculating time shortage can make efficiency reduce.And, the risk losing effective information can cause one relatively The signal to noise ratio result of difference.But, when Decomposition order deficiency, optimum can not be obtained equally.Therefore, The present invention selects three layers of decomposition.

(2) denoising effect

The line being located below in Fig. 3 represents primary signal, and another represents and uses based on signal intensity self adaptation Lifting Wavelet dual threshold Denoising Algorithm (present invention) that TABU search optimizes carries out the signal after denoising.Permissible It is evident that along with the raising of input signal power, not only the output signal-to-noise ratio of primary signal improves, denoising The output signal-to-noise ratio of signal improves the most therewith.And, the difference between them is held essentially constant, this meaning The system of wearing can process this input signal power small-signal between-176dBw to-168dBw, at letter Make an uproar than the gain that above can provide 8dB.

(3) present invention and the contrast of other algorithm

Fig. 4 shows the comparison diagram of output signal-to-noise ratio in the case of four kinds.Line with round dot represents this The denoising effect of the algorithm of bright middle proposition, the Lifting Wavelet i.e. optimized based on signal intensity self adaptation TABU search Dual threshold Denoising Algorithm.It is obvious that fact proved, it is possible to the method obtaining optimal output signal-to-noise ratio is this The algorithm of bright middle proposition, and be used alone Lifting Wavelet carry out the effect of denoising be better than be used alone tradition from Dissipate wavelet transformation (DWT).Meanwhile, along with the increase of input signal power, it is used alone DWT and individually makes The snr gain being obtained in that with Lifting Wavelet denoising is gradually lowered, on the contrary, adaptive based on signal intensity The performance answering the Lifting Wavelet dual threshold Denoising Algorithm of TABU search optimization is stable and gratifying.

(4) application wavelet decomposition lifting scheme and the contrast on time complexity of the DWT scheme

In the following table, compared for the calculating time of two kinds of denoising methods, i.e. avoiding based on signal intensity self adaptation The DWT small echo dual threshold Denoising Algorithm of chess game optimization and based on signal intensity self adaptation TABU search optimization carry Rise small echo dual threshold Denoising Algorithm.Should be noted that, the specific calculating time can not represent real reality Time, but this contrast can be regarded as computation complexity and the reflection of enforcement difficulty.Therefore, it can Reaching a conclusion, DWT needs to consume the time far above application wavelet decomposition lifting scheme.

It is emphasized that embodiment of the present invention is illustrative rather than determinate, therefore originally Invention includes the embodiment that is not limited to described in detailed description of the invention, every by those skilled in the art according to Other embodiments that technical scheme draws, also belong to the scope of protection of the invention.

Claims (5)

1. a Lifting Wavelet dual threshold Denoising Algorithm based on signal intensity self adaptation TABU search, its feature It is to comprise the following steps:

Step 1, utilize differential coherence Cumulate algorithm capture faint received global navigation satellite system signal；

Faint received global navigation satellite system signal is analyzed by step 2, application Lifting Wavelet decomposition；

Dual threshold in Lifting Wavelet denoising is selected by step 3, application signal intensity self adaptation tabu search algorithm It is optimized；

Step 4, utilize optimized dual threshold to capture faint received global navigation satellite system signal carry out after locate Reason.

Lifting Wavelet dual threshold based on signal intensity self adaptation TABU search the most according to claim 1 Denoising Algorithm, it is characterised in that: the implementation method of described step 1 is: obtain micro-by differential coherence accumulation Weak signal, then inputs Wavelet Denoising Method wave filter to find its code phase and Doppler frequency shift by accumulation results, Thus complete acquisition procedure.

Lifting Wavelet dual threshold based on signal intensity self adaptation TABU search the most according to claim 1 Denoising Algorithm, it is characterised in that: the concrete step that processes of described step 2 is:

(1) decomposition step, by original signal sequence S_jIt is decomposed into two subdivisions according to parity even_j-1And odd_j-1:

(odd_j-1,even_j-1)=Split (S_j)

(2) prediction steps, is predicted by following prediction machine:

d_j-1=odd_j-1-P(even_j-1)

Wherein, P (even_j-1) it is to predict odd number value, specifying information d by even item_j-1Pass through odd_j-1With in advance Survey P (even_j-1Between), error represents；

(3) update step, update average as the following formula:

S_j-1=even_j-1+U(d_j-1)。

Lifting Wavelet dual threshold based on signal intensity self adaptation TABU search the most according to claim 1 Denoising Algorithm, it is characterised in that: the concrete step that processes of described step 3 is:

(1) upper threshold and lower limit T are supposed_m'_inAnd T_m'_ax

(2) the initial interval of threshold optimization is set to [T_min,T_max], by upper limit T_maxIt is set to T_max=3 σ,

Wherein σ is estimated as: σ ≈ Mid/0.6745, in formula, Mid be Decomposition order minimum time wavelet coefficient by Median after numerical ordering, lower limit T_minIt it is the absolute value of minimum coefficient than 0 in jth layer wavelet decomposition；

(3) at T '_minWith T '_maxBetween produce a candidate solution subset, then according to following fitness function It is ranked up:

$S=\mathrm{\&alpha;}\&times;\lg \&lsqb;\frac{\underset{n}{\&Sigma;}{f}_d^2\left(n\right)}{\underset{n}{\&Sigma;}{\&lsqb;f_d\left(n\right)-f\left(n\right)\&rsqb;}^2}\&rsqb;$

Wherein, α is the auto-adaptive parameter that the strength information according to faint received global navigation satellite system signal is chosen, f_dFor Reconstruction signal after denoising, f is reference signal；

$\mathrm{\&alpha;}=\sqrt{{\mathrm{\&sigma;}}_{\mathrm{\&omega;}}^2\left(t\right)+{\mathrm{\&sigma;}}_n^2}$

Wherein, σ_ωT () is the root-mean-square value of signal,For system average noise power；

(4) scan for by signal intensity self adaptation TS algorithm, be modified current solution obtaining feasible Solving, if moving to its neighbouring solution to be not better than the most current optimal solution, the most adjacent solution will be accepted；Then draw Enter to avoid list and avoid circulation；

(5) step (3) is returned until meeting stopping criterion；

(6) utilize equation below calculate optimal threshold:

T′_max=y_1j×10⁰+y_2j×10^-1+y_3j×10^-2+y_4j×10^-3

T′_min=y_5j×10⁰+y_6j×10^-1+y_7j×10^-2+y_8j×10^-3。

Lifting Wavelet dual threshold based on signal intensity self adaptation TABU search the most according to claim 1 Denoising Algorithm, it is characterised in that: the concrete processing method of described step 4 is: at optimal double threshold value [T '_min,T′_max] After being determined, kth wavelet coefficient W in jth layer wavelet decomposition_j,kProcess as the following formula:

$W_{j,k}^{\&prime;}=\left\{\begin{array}{c}0,\left|W_{j,k}\right|<T_{\min }^{\&prime;};\\ \frac{T_{\max }^{\&prime;}}{T_{\max }^{\&prime;}-T_{\min }^{\&prime;}}\left(\right|W_{j,k}|-T_{\min }^{\&prime;})\mathrm{sgn}\left(W_{j,k}\right),T_{\min }^{\&prime;}\&le;\left|W_{j,k}\right|\&le;T_{\max }^{\&prime;}.\\ W_{j,k},\left|W_{j,k}\right|>T_{\max }^{\&prime;}\&CenterDot;\end{array}\right.$