Impulsive Noise Mitigation Method in a kind of power line communication based on compressed sensing
Technical field
Impulsive Noise Mitigation Method in being communicated the present invention relates to one kind, more particularly, to a kind of electric power based on compressed sensing
Impulsive Noise Mitigation Method in line communication.
Background technology
Compared with other communication technologys, PLC (Power Line Communications, power line communication) technology is applicable
Property is wide, and installation fee is low, and this has made it communication mode very popular in small grids field and other application.But PLC
The application of technology receives the limitation of some unfavorable factors, wherein impulsive noise be influence data on electric network transmission it is main because
Element.Impulsive noise is roughly divided into two types:Asynchronous type and preiodic type.Asynchronous type impulsive noise is mainly the switch by appliance chamber
Caused by transient state, it is characterized in that the duration is short, pulse power is high, random to occur.OFDM (Orthogonal
Frequency Division Multiplexing, OFDM) technology can effectively tackle frequency selective fading channels, institute
It is more much smaller than single carrier with the susceptibility to impulsive noise.Therefore, OFDM is obtained in the substandard physical layers of newest PLC
It is widely applied.
The common methods Finite Amplitude of existing asynchronous pulse noise noise reduction and blanking.But both approaches will make an uproar to pulse
Sound parameter or ofdm signal parameter are estimated.But in a plc environment, impulsive noise changes at any time, therefore to parameter
Accurate measurement is more difficult.It is openness in time domain in view of asynchronous pulse noise, i.e. umber of pulse in an OFDM symbol
It is applied to pulse noise reduction not over some threshold values, thus by compressed sensing (Compressed Sensing, CS) technology.Pulse
The openness of noise also determines that it can be calculated by management loading (Sparse Bayesian Learning, SBL)
Method is restored.On this basis, it is thus proposed that two kinds of improved SBL algorithms:A kind of is the calculation using whole carrier informations
Method (SBL with all tones);Another kind is judgment feed back type SBL algorithms (SBL with DF).These SBL algorithms pass through
The prior information for closing impulsive noise is integrated with, improves the effect and intensity of noise reduction well, but algorithm has higher meter
Calculate complexity.
The content of the invention
The technical problems to be solved by the invention be to provide that a kind of computation complexity is relatively low and excellent noise reduction effect based on pressure
Impulsive Noise Mitigation Method in the power line communication perceived that contracts.
Technical scheme is used by the present invention solves above-mentioned technical problem:A kind of power line communication based on compressed sensing
Middle Impulsive Noise Mitigation Method, it is characterised in that null subcarrier and pilot tone in the OFDM symbol received first with receiving terminal
Subcarrier information, an impulsive noise restructing algorithm framework is constructed using compressed sensing algorithm, in receiving terminal, believed using transmitting
Number algebraic characteristic keep constant this feature, a constraint bar is introduced under the impulsive noise restructing algorithm framework of structure
Part, by convex relaxation, using the rectangular extent loose constraint condition in modulation constellation, using a variety of existing tool boxes, lead to
The estimate that impulsive noise is calculated is crossed, and by subtracting the impulsive noise estimate, complete paired pulse in reception signal
The elimination of noise, is comprised the following steps that:
1. in transmitting terminal, frequency-domain OFDM symbol is passed through into x=F*Time-domain OFDM symbol x is converted into after X mappings, wherein, X is
Frequency-domain OFDM symbol, F are leaf transformation matrix, F in N point discrete Fouriers*It is F Hermitian transformation, then, length is previously inserted into x
The cyclic prefix extended more than the maximum delay of channel;
2. the time domain OFDM signal that receiver is received is defined as r=Hx+i+g, wherein, H is N × N circular matrixes, N
It is the sum of OFDM symbol subcarrier, H first row is made up of normalized channel impulse response, and i represents impulsive noise vector,
G represents additive Gaussian ambient noise (Additive White Gaussian Noise, AWGN) vector, and r, x, i and g are set to
In multidimensional complex signal space CNIn;
3. by ofdm demodulator, the time domain OFDM signal r received is converted to by Discrete Fourier transform
Frequency domain ofdm signal Y:Y=Fr=FHF*X+Fi+Fg=Λ X+Fi+G, wherein Λ=FHF*=diag (H0,H1,...,HN-1) be
One diagonal matrix, represent imperfect channel state information, its diagonal entry H0,H1,…,HN-1Corresponding to channel impulse response
Leaf transformation matrix coefficient in N point discrete Fouriers;G=Fg is g Discrete Fourier transform;
4. setting indexed sets of the D as non-data subcarrier, its sum is | D |=M, define YD、ΛD、XD、FDAnd GDRespectively
It is the part for corresponding to non-data subcarrier in Y, Λ, X, F and G, obtains the OFDM symbols that non-data received over subcarriers arrives
Number YD=ΛDXD+FDi+GD=(Λ X)D+FDi+GD, define one group of Systems with Linear Observation data of impulsive noiseAnd it is sparse in time domain to define impulsive noise, using in compression sensing method
Minimum l1- norm restructing algorithm, impulsive noise estimation formulas is converted into the convex optimization problem of following Second-order cone programming:Wherein ε1>0 is the maximum allowable root-mean-square error of ambient noise, takes ε1=10-3, | | | |1With
||·||2The l of vector is represented respectively1- norm and l2- norm;
5. setIt is set D supplementary set, represents the indexed set of data subcarrier, thenRepresent that data carries
The number of ripple, according to step 4., it is defined on the OFDM symbol on real data subcarrier in the presence of ambient noiseNoise
Form beWherein,WithBe respectively Λ,
Y, the part of data subcarrier is corresponded in F, X and G,Represent the AWGN vectors on data subcarrier, ()-1Representing matrix
Inverse operation, utilize testing conditions of the data subcarrier in modulation constellation, foundationOFDM symbol after equilibriumWithJudgement symbol in modulation constellationBetween Euclidean distance it is minimum, one new optimization problem of construction is such as
Under:
min||i||1
s.t.||z-FDi||2≤ε1
Wherein,The element in data subcarrier indexed set is represented, Re () represents to take real part computing, Im () table
Show and take imaginary-part operation, ε2For a given less value, ε is taken2=10-4;
6. carrying out the approximation of linear programming problem to the optimization problem of step 5., if Ω represents the set of constellation point, Ω is usedRe
=Re (Ω) and ΩIm=Im (Ω) represents the set of the real coordinate of all constellation points and empty coordinate respectively, according to the symmetrical of Ω
Property, definition planisphere are a rectangular area, and its border is given by
Wherein uRRepresent the length of rectangular area, uIThe width of rectangular area is represented, when the planisphere of modulation is square, uR=uI, then will
Step 5. in optimization problem constraints slacking beBy step 5. in it is excellent
Change problem is rewritten as convex problemAccording to the ofdm signal of time domain receiving terminal,
The summation of impulsive noise and ambient noise is expressed as e=i+g, then makes z=FDE, will | | e | |1Regard as and have ignored ambient noise
After g | | i | |1Approximation, then use | | e | |1Replace | | i | |1, obtain following linear programming problem:
7. the linear programming problem 6. obtained by solution procedure, obtain the estimate of impulsive noiseThen connecing
The estimation of impulsive noise is subtracted in the collection of letters number, completes the elimination of impulsive noise.
Compared with prior art, the advantage of the invention is that:
1. the present invention utilizes non-data carrier information and compressed sensing technology, impulsive noise estimation problem is converted into l1-
Norm minimum problem, and classical Second-order cone programming method for solving is utilized, as CVX tool boxes are solved.
2. the problem of causing throughput of system to lose for increase non-data carrier wave, the present invention proposes to believe using data carrier
Lifting system performance is ceased, and using Euclidean distance during modulation constellation judgement as constraints, constructing one has about
The minimum impulsive noise l of beam condition1- norm problem is estimated impulsive noise, adds the degree of accuracy of estimation, improves
Systematic function.Simulation result is shown, in the uncoded system of 4QAM modulation, when bit error rate is 10-3When, compared to its other party
Method, signal to noise ratio improve 3-8dB;In the coded system of 4QAM modulation, when bit error rate is 10-4When, compared to other method,
Signal to noise ratio improves 4-10dB.In the uncoded system of 16QAM modulation, when bit error rate is 10-3When, compared to its other party
Method, signal to noise ratio improve 4-8dB;In the coded system of 16QAM modulation, when bit error rate is 10-3When, compared to its other party
Method, signal to noise ratio improve 7-10dB.
3. for the minimum impulsive noise l with constraints using data carrier information structuring1- norm problem exists
Np hard problem under high order modulation, using convex relaxation method by former constraints slacking into a rectangular area, so that non-
Convex constraint is converted into convex constraint, and is solved using linear programming method, reduces the complexity of solution, allows it more
Complete to solve in item formula time complexity.
Brief description of the drawings
Fig. 1 is PLC system module frame chart;
Fig. 2 is the FB(flow block) of the inventive method;
Fig. 3 is the various performance of BER comparison schematic diagrams of the method for eliminating audible noise in uncoded 4-QAM systems;
Fig. 4 is the various performance of BER comparison schematic diagrams of the method for eliminating audible noise in 4-QAM systems are encoded;
Fig. 5 is the various performance of BER comparison schematic diagrams of the method for eliminating audible noise in uncoded 16-QAM systems;
Fig. 6 is the various performance of BER comparison schematic diagrams of the method for eliminating audible noise in 16-QAM systems are encoded.
Embodiment
The present invention is described in further detail below in conjunction with accompanying drawing embodiment.
As shown in Fig. 2 Impulsive Noise Mitigation Method in a kind of power line communication based on compressed sensing, first with reception
The null subcarrier and pilot sub-carrier information in the OFDM symbol received are terminated, a pulse is constructed using compressed sensing algorithm
Noise restructing algorithm framework, in receiving terminal, this constant feature is kept using the algebraic characteristic of transmission signal, in the pulse of structure
A constraints is introduced under noise restructing algorithm framework, by convex relaxation, utilizes the rectangular extent pine in modulation constellation
Relaxation constraints, estimate by the way that the estimate of impulsive noise is calculated, and by subtracting the impulsive noise in reception signal
Value, completes the elimination to impulsive noise, comprises the following steps that:
1. in transmitting terminal, frequency-domain OFDM symbol is passed through into x=F*Time-domain OFDM symbol x is converted into after X mappings, wherein, X is
Frequency-domain OFDM symbol, F are leaf transformation matrix, F in N point discrete Fouriers*It is F Hermitian transformation, then, length is previously inserted into x
The cyclic prefix extended more than the maximum delay of channel;
2. the time domain OFDM signal that receiver is received is defined as r=Hx+i+g, wherein, H is N × N circular matrixes, N
It is the sum of OFDM symbol subcarrier, H first row is made up of normalized channel impulse response, and i represents impulsive noise vector,
G represents additive Gaussian ambient noise (Additive White Gaussian Noise, AWGN) vector, and r, x, i and g are set to
In multidimensional complex signal space CNIn;
3. by ofdm demodulator, the time domain OFDM signal r received is converted to by Discrete Fourier transform
Frequency domain ofdm signal Y:Y=Fr=FHF*X+Fi+Fg=Λ X+Fi+G, wherein Λ=FHF*=diag (H0,H1,...,HN-1) be
One diagonal matrix, represent imperfect channel state information, its diagonal entry H0,H1,…,HN-1Corresponding to channel impulse response
Leaf transformation matrix coefficient in N point discrete Fouriers;G=Fg is g Discrete Fourier transform;
4. setting indexed sets of the D as non-data subcarrier, its sum is | D |=M, define YD、ΛD、XD、FDAnd GDRespectively
It is the part for corresponding to non-data subcarrier in Y, Λ, X, F and G, obtains the OFDM symbols that non-data received over subcarriers arrives
Number YD=ΛDXD+FDi+GD=(Λ X)D+FDi+GD, define one group of Systems with Linear Observation data of impulsive noiseAnd it is sparse in time domain to define impulsive noise, using in compression sensing method
Minimum l1- norm restructing algorithm, impulsive noise estimation formulas is converted into the convex optimization problem of following Second-order cone programming:Wherein ε1>0 is the maximum allowable root-mean-square error of ambient noise, takes ε1=10-3, | | | |1With
||·||2The l of vector is represented respectively1- norm and l2- norm;
5. setIt is set D supplementary set, represents the indexed set of data subcarrier, thenRepresent that data carries
The number of ripple, according to step 4., it is defined on the OFDM symbol on real data subcarrier in the presence of ambient noiseNoise
Form beWherein,WithBe respectively Λ,
Y, the part of data subcarrier is corresponded in F, X and G,Represent the AWGN vectors on data subcarrier, ()-1Representing matrix
Inverse operation, utilize testing conditions of the data subcarrier in modulation constellation, foundationOFDM symbol after equilibriumWithJudgement symbol in modulation constellationBetween Euclidean distance it is minimum, one new optimization problem of construction is such as
Under:
min||i||1
s.t.||z-FDi||2≤ε1
Wherein,The element in data subcarrier indexed set is represented, Re () represents to take real part computing, Im () table
Show and take imaginary-part operation, ε2For a given less value, ε is taken2=10-4;
6. carrying out the approximation of linear programming problem to the optimization problem of step 5., if Ω represents the set of constellation point, Ω is usedRe
=Re (Ω) and ΩIm=Im (Ω) represents the set of the real coordinate of all constellation points and empty coordinate respectively, according to the symmetrical of Ω
Property, definition planisphere are a rectangular area, and its border is given by
Wherein uRRepresent the length of rectangular area, uIThe width of rectangular area is represented, when the planisphere of modulation is square, uR=uI, then will
Step 5. in optimization problem constraints slacking beBy step 5. in it is excellent
Change problem is rewritten as convex problemAccording to the ofdm signal of time domain receiving terminal,
The summation of impulsive noise and ambient noise is expressed as e=i+g, then makes z=FDE, will | | e | |1Regard as and have ignored ambient noise
After g | | i | |1Approximation, then use | | e | |1Replace | | i | |1, obtain following linear programming problem:
7. the linear programming problem 6. obtained by solution procedure, obtain the estimate of impulsive noiseThen connecing
The estimation of impulsive noise is subtracted in the collection of letters number, completes the elimination of impulsive noise.
In order to further illustrate, the Computer simulation results of proposed pulse noise-reduction method are presented herein.Simulation is
Carried out in the complicated base band of the PLC system based on OFDM, impulsive noise model uses Gaussian Mixture (Gaussian
Mixture, GM) model and Myddelton A classes (Middleton Class A, MCA) two kinds of model, simulation parameter sets such as table
Shown in lattice I.
Form I. analog parameters
Under identical simulated environment, we compare method of the invention and some existing noise-reduction methods in errored bit
Performance in terms of rate (Bit Error Rate, BER).In following chart, method proposed by the invention is denoted as " CS
with all tones”.The OFDM receiver for not doing impulsive noise elimination is denoted as " No Mitigation ", other several bases
" CS ", " SBL with null tones " and " SBL with all tones " are respectively designated as in the method for compressed sensing.For
Make ratio between nonparametric and parametric technique, we also present lowest mean square root error (Minimum Mean Square
Error, MMSE) detector result.Two kinds of situations of the presence or absence of MMSE detectors noise states information are respectively designated as " MMSE
W/NSI " and " MMSE W/O NSI ".It is noted that " MMSE w/NSI " detections are the best approaches in parametric technique.
Fig. 3 to Fig. 6 is compared using in the system of the uncoded of 4-QAM and 16-QAM and coding, and above-mentioned all pulses are made an uproar
The BER performances of sound suppressing method.There is no " the analog result of SBL with DF " methods, because this method needs in Fig. 3 and Fig. 5
The prior information of the covariance matrix of impulsive noise is wanted, therefore is not suitable for uncoded system.
In figure 3, we analyze the result of uncoded 4-QAM systems first.Obviously, except " MMSE with NSI "
This situation that detector is put up the best performance in relatively low signal to noise ratio region, our method are better than every other method.But with
The raising of signal to noise ratio, our method starts in higher signal to noise ratio (Signal to Noise Ratio, SNR) region
Show big advantage.When only considering the impulsive noise estimation on non-data carrier wave, in two kinds of noise models, original CS
Algorithm is than " snr gain that SBL with null tones " algorithms are obtained is few.On the contrary, examined when whole carrier waves are included
During worry, proposed by the present invention " CS with all tones " methods show the advantage more than SBL algorithms, in GM models
Snr gain is more than 2dB, in MCA models, then in 4dB between 6dB.
Then, we analyze the result of the coded system using 4-QAM in Fig. 4.Contrast uncoded system in figure 3
System, method proposed by the present invention show the performance better than most of algorithms again.Compared with DF SBL algorithms, present invention side
Method has faint drop in the low signal-to-noise ratio region of GM impulsive noises.But under MCA models, the inventive method obtains
One relatively large snr gain.
What Fig. 5 and Fig. 6 was presented is the analog result under 16-QAM, it can be seen that the BER performances of all noise-reduction methods are not
All reduced in the case of coding and coding.This can become more from the constellation point of high order modulation to given transmission power condition
Harsh to explain, also because of that, the detection of OFDM data symbol is easier to be disturbed by impulsive noise residual.Its result
It is that on larger SNR value, " MMSE with NSI " performance becomes than " the SBL with null in uncoded system
Tones " and " SBL with DF " are worse.And " SBL with null tones " become than " SBL with all tones "
More effectively.Through observation shows that in 16-QAM systems, the inventive method can be obtained than its other party in larger SNR regions
The all more significant performance gain of method, therefore the inventive method has preferable performance.