CN104079520A - Impulse interference inhibition method of OFDM system - Google Patents

Abstract

The invention discloses an impulse interference inhibition method of an OFDM system. According to the method, solving the problem for calculating the estimation value of an unknown impulse noise vector is converted into solving the problem enabling the value of a l2-l1 norm type objective function to be the smallest, an optimization sub-problem is solved through the iteration mode to obtain the optimal solution enabling the value of the l2-l1 norm type objective function to be the smallest, and a receipt signal vector after the pulse noise interference is inhibited is obtained by subtracting the final estimation value of an impulse noise vector from the receipt signal vector with a cyclic prefix removed and the discrete Fourier transform implemented through the discrete Fourier transform. The method has the advantages that solving the original problem is converted into solving the problem enabling the value of the l2-l1 norm type objective function to be the smallest, and the NP-hard problem in the process of directly estimating the optimal sparse solution of the impulse noise can be avoided; the optimal solution enabling the value of the l2-l1 norm type objective function to be the smallest is obtained by solving the optimization sub-problem in the iteration mode, so that the time used for estimating the pulse noise is shortened, and the influence on the OFDM symbol error probability from the pulse noise is reduced.

Description

A kind of pulse interference suppression method of ofdm system

Technical field

The present invention relates to a kind of pulse interference suppression technology of the communications field, especially relate to a kind of pulse interference suppression method of ofdm system.

Background technology

Due to OFDM (Orthogonal Frequency Division Multiplexing, OFDM) technology has good anti-multipath interference performance, therefore it is adopted by standards such as IEEE802.11, IEEE802.16 and IEEE802.20, and is widely used in various digital broadband communication systems.But, in many practical communication system, when an ofdm signal is after transmission, receiving the impulsive noise that has not only comprised Gaussian Background noise in signal but also also had many artificial generations, these impulsive noises are completely different from the statistical property of Gaussian Background noise, cause the performance of traditional ofdm system based on background noise design under impulse noise interference environment seriously to decay.Therefore, designing effective Impulse Noise Interference is very necessary for the performance that promotes ofdm system.

Studied widely at present about the impulse noise mitigation problem in ofdm system.The people such as Himal A.Suraweera carry out by increase a kind of nonlinear filter that presets decision threshold before traditional OFDM receiver the impulsive noise that filtering amplitude is larger, have realized the inhibition of paired pulses noise jamming.The people such as Pablo Torio have proposed a kind of ofdm system impulse noise suppression method interweaving based on cell.In addition, the people such as Takuya Kitamura have studied the Impulse Noise Interference based on copy signal, carry out symbol judgement and copy signal is deleted the elimination that has realized paired pulses noise by iteration.But, more than research all supposes that impulsive noise is to obey a certain specific noise statistics distributed model, this causes in the time processing with the signal of impulse noise interference, need first to obtain concrete impulsive noise statistical distribution parameter through a data-aided training process, not only increase system complexity, and in the time that the noise statistics distributed model of supposition can not reflect actual impulsive noise statistical property, the performance of these class methods can serious decay.For this problem, the people such as Lin and Evans utilizes the sparse characteristic of impulsive noise in time domain, a kind of Impulsive Noise Mitigation Method based on compressed sensing technology has been proposed, the method does not need to preset the statistical model of impulsive noise, and can learn by sparse Bayesian (Sparse Bayesian learning, SBL) restructing algorithm estimates the impulsive noise vector of higher-dimension from the known signal vector compared with low-dimensional, but because SBL algorithm is in the time solving super parameter by maximum likelihood method, directly utilize conditional expectation to maximize the super parameter of simultaneously upgrading need estimation in each iterative process, this makes whole SBL convergence of algorithm slower, and can cause conditional expectation maximization steps to be difficult to realize, and then make the time complexity of whole method higher, practicality is not strong.

Summary of the invention

Technical problem to be solved by this invention is to provide a kind of pulse interference suppression method of ofdm system, and it can reduce the time of estimating for impulsive noise effectively, and can effectively reduce the impact of impulsive noise on OFDM symbol error probability.

The present invention solves the problems of the technologies described above adopted technical scheme: a kind of pulse interference suppression method of ofdm system, is characterized in that comprising the following steps:

1. at transmitting terminal, ofdm signal vector to be sent is designated as to X, supposes to have M known pilot in X, the vector M in X known pilot being formed is designated as X _t; Then X successively carried out inverse discrete Fourier transformer inverse-discrete and adds Cyclic Prefix, obtaining transmission signal vector; Wherein, the dimension of X is N × 1 dimension, N>2,1<M<N, X _tdimension be M × 1 dimension, T represents the indexed set being made up of the index of the known pilot of the M in X, the length of T is M;

2. supposition is known for the state information of the channel of signal transmission, transmitting terminal passes through transmission transmission signal vector to receiving terminal, the impulsive noise vector sum background noise vector correspondence that transmission signal vector is produced in transmitting procedure in channel is designated as e and n, wherein, the dimension of unknown impulsive noise vector e and unknown background noise vector n is N × 1 dimension;

3. at receiving terminal, the received signal vector receiving is successively removed to Cyclic Prefix and discrete Fourier transform (DFT), obtain pending received signal vector, be designated as Y, Y forms after by X superimposed pulse noise vector e and background noise vector n; Then from Y, extract at X _tthe signal vector forming after middle superposeed impulsive noise vector e and background noise vector n, is designated as Y _t; According to the state information of channel, remove Y again _tin with X _tcorresponding known signal vector, is only comprised the known vector Z of unknown impulsive noise vector e and unknown two compositions of background noise vector n; Wherein, the dimension of Y is N × 1 dimension, Y _tdimension be M × 1 dimension, the dimension of Z is M × 1 dimension;

4. the estimated value problem of asking for unknown impulsive noise vector e is converted into the estimated value that solves an e and makes l ₂-l ₁norm type target function the problem of value minimum, then will in be expressed as f (e) with functional form, wherein, symbol " || || ₂" for asking for l ₂norm symbol, symbol for asking for l ₂norm square, symbol " || || ₁" for asking for l ₁norm symbol, F _tthe discrete Fourier transform (DFT) matrix that represents known M × N dimension, τ represents regularization parameter, f () is function representation mode;

5. the estimated value that solves an e by iterative manner makes l ₂-l ₁norm type target function value minimum, detailed process is:

5.-1, the value of setting regularization parameter τ is τ=0.1|| (F _t) ^hz|| _∞, the initial value that makes the estimated value of e is e ₀, e ₀=((F _t) ^hf _t) ^-1(F _t) ^hz, and make k represent iterations, the initial value of k is 1, wherein, symbol " || || _∞" for asking for Infinite Norm symbol, (F _t) ^hfor F _tassociate matrix, ((F _t) ^hf _t) ^-1for (F _t) ^hf _tinverse matrix;

5.-2, the evaluation factor-alpha of the k time iteration of definition _k, ${\mathrm{\&alpha;}}_k=\left\{\begin{array}{cc}1& k=1\\ \frac{{\left(S_k\right)}^H{r}_k}{{\left(S_k\right)}^H{S}_k}& k\&NotEqual;1\end{array}\right.,$ Make again wherein, (S _k) ^hfor S _kassociate matrix, S _k=e _k-e _k-1, e _krepresent the estimated value of the e obtaining after the k time iteration, e _k-1represent the estimated value of the e obtaining after the k-1 time iteration, represent to ask for f (e after the k time iteration _k) first derivative values, represent to ask for f (e after the k-1 time iteration _k-1) first derivative values,

$f\left(e_{k-1}\right)=\frac{1}{2}{\left|\right|Z-F_T{e}_{k-1}\left|\right|}_2^2;$

5.-3, according to α _kand u _k, calculate the estimated value of the e obtaining after the k time iteration, be designated as e _k, wherein, max () is for getting max function, symbol " | | " be the symbol that takes absolute value, represent u _kin i element, 1≤i≤N;

5.-4, judgement whether set up, if set up, execution step 5.-5; Otherwise, make k=k+1, then return to step 5.-2 and continue to carry out, wherein, symbol " || || ₂" for asking for l ₂norm symbol, ε is the minimum normal number of setting, "=" in k=k+1 is assignment;

5.-5, finishing iteration process, obtains the final estimated value of e, is designated as wherein, in "=" be assignment;

6. deduct the final estimated value of e with Y the frequency-domain expression obtaining by discrete Fourier transform (DFT), the received signal vector being inhibited after impulse noise interference.

In described step 5.-4, get ε=10 ^-5.

Compared with prior art, the invention has the advantages that:

1) the inventive method is converted into the estimated value problem of asking for unknown impulsive noise vector to solve one and make l ₂-l ₁the problem of the value minimum of norm type target function, the NP-hard problem occurring can effectively avoid so the optimum sparse solution of direct estimation impulsive noise time.

2) the inventive method solves one by iterative manner and optimizes subproblem and obtain one and make l ₂-l ₁the optimal solution of the value minimum of norm type target function, has not only reduced significantly the time of estimating for impulsive noise, and has effectively reduced the impact of impulsive noise on OFDM symbol error probability, has so just improved practicality.

Brief description of the drawings

Fig. 1 adopts the inventive method to realize the system block diagram of pulse interference suppression;

Fig. 2 is signal to noise ratio one timing, adopts the contrast schematic diagram of the average operating time of the CPU that the inventive method and existing impulse noise suppression method based on SBL consume in the time carrying out ofdm system impulse noise mitigation;

Fig. 3 is the comparison schematic diagram of the inventive method and existing impulse noise suppression method based on the SBL symbol error probability of ofdm system while changing with signal to noise ratio.

Embodiment

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

The pulse interference suppression method of a kind of ofdm system that the present invention proposes, Fig. 1 has provided and has adopted the inventive method to realize the system block diagram of pulse interference suppression, and the inventive method specifically comprises the following steps:

1. at transmitting terminal, ofdm signal vector to be sent is designated as to X, supposes to have M known pilot in X, the vector M in X known pilot being formed is designated as X _t; Then X successively carried out inverse discrete Fourier transformer inverse-discrete and adds Cyclic Prefix, obtaining transmission signal vector; Wherein, the dimension of X is N × 1 dimension, N>2,1<M<N, X _tdimension be M × 1 dimension, T represents the indexed set being made up of the index of the known pilot of the M in X, the length of T is M.

In the specific implementation, N generally can value be 16,64,128,512 or 1024 etc., and M generally gets the half of N, $M=\frac{N}{2}.$

2. supposition is known for the state information of the channel of signal transmission, transmitting terminal passes through transmission transmission signal vector to receiving terminal, the impulsive noise vector sum background noise vector correspondence that transmission signal vector is produced in transmitting procedure in channel is designated as e and n, wherein, the dimension of unknown impulsive noise vector e and unknown background noise vector n is N × 1 dimension.

3. at receiving terminal, the received signal vector receiving is successively removed to Cyclic Prefix and discrete Fourier transform (DFT), obtain pending received signal vector, be designated as Y, Y forms after by X superimposed pulse noise vector e and background noise vector n; Then from Y, extract at X _tthe signal vector forming after middle superposeed impulsive noise vector e and background noise vector n, is designated as Y _t; According to the state information of channel, remove Y again _tin with X _tcorresponding known signal vector, is only comprised the known vector Z of unknown impulsive noise vector e and unknown two compositions of background noise vector n, vector Z can be called to the projection of unknown impulsive noise vector in known pilot at this; Wherein, the dimension of Y is N × 1 dimension, Y _tdimension be M × 1 dimension, the dimension of Z is M × 1 dimension.

4. the estimated value problem of asking for unknown impulsive noise vector e is converted into the estimated value that solves an e and makes l ₂-l ₁norm type target function the problem of value minimum, then will in be expressed as f (e) with functional form, wherein, symbol " || || ₂" for asking for l ₂norm symbol, symbol for asking for l ₂norm square, symbol " || || ₁" for asking for l ₁norm symbol, F _tthe discrete Fourier transform (DFT) matrix that represents known M × N dimension, τ represents regularization parameter, and the value of τ is the normal number of setting, f () is function representation mode.

5. solving one by iterative manner optimizes the estimated value that subproblem obtains an e and makes l ₂-l ₁norm type target function value minimum, detailed process is:

5.-1, the value of setting regularization parameter τ is τ=0.1|| (F _t) ^hz|| _∞, the initial value that makes the estimated value of e is e ₀, e ₀=((F _t) ^hf _t) ^-1(F _t) ^hz, and make k represent iterations, the initial value of k is 1, wherein, symbol " || || _∞" for asking for Infinite Norm symbol, (F _t) ^hfor F _tassociate matrix, ((F _t) ^hf _t) ^-1for (F _t) ^hf _tinverse matrix.

5.-2, the evaluation factor-alpha of the k time iteration of definition _k, ${\mathrm{\&alpha;}}_k=\left\{\begin{array}{cc}1& k=1\\ \frac{{\left(S_k\right)}^H{r}_k}{{\left(S_k\right)}^H{S}_k}& k\&NotEqual;1\end{array}\right.,$ Make again wherein, (S _k) ^hfor S _kassociate matrix, S _k=e _k-e _k-1, e _krepresent the estimated value of the e obtaining after the k time iteration, e _k-1represent the estimated value of the e obtaining after the k-1 time iteration, represent to ask for f (e after the k time iteration _k) first derivative values, represent to ask for f (e after the k-1 time iteration _k-1) first derivative values,

$f\left(e_{k-1}\right)=\frac{1}{2}{\left|\right|Z-F_T{e}_{k-1}\left|\right|}_2^2.$

5.-3, according to α _kand u _k, calculate the estimated value of the e obtaining after the k time iteration, be designated as e _k, $e_k=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\frac{\max \left(\right|u_k^i|-\frac{\mathrm{\&tau;}}{{\mathrm{\&alpha;}}_k},0)}{\max \left(\right|u_k^i|-\frac{\mathrm{\&tau;}}{{\mathrm{\&alpha;}}_k},0)+\frac{\mathrm{\&tau;}}{{\mathrm{\&alpha;}}_k}}{u}_k^i,$ Wherein, will at this $\frac{\max \left(\right|u_k^i|-\frac{\mathrm{\&tau;}}{{\mathrm{\&alpha;}}_k},0)}{\max \left(\right|u_k^i|-\frac{\mathrm{\&tau;}}{{\mathrm{\&alpha;}}_k},0)+\frac{\mathrm{\&tau;}}{{\mathrm{\&alpha;}}_k}}{u}_k^i$ Be called and optimize subproblem, max () is for getting max function, symbol " | | " be the symbol that takes absolute value, represent u _kin i element, 1≤i≤N.

5.-4, judgement whether set up, if set up, execution step 5.-5; Otherwise, make k=k+1, then return to step 5.-2 and continue to carry out, wherein, symbol " || || ₂" for asking for l ₂norm symbol, ε is the minimum normal number of setting, and gets in the present embodiment ε=10 ^-5, "=" in k=k+1 is assignment.

5.-5, finishing iteration process, obtains the final estimated value of e, is designated as wherein, in "=" be assignment.

6. deduct the final estimated value of e with Y the frequency-domain expression obtaining by discrete Fourier transform (DFT), the received signal vector being inhibited after impulse noise interference.

By following emulation to further illustrate feasibility and the validity of pulse interference suppression method of ofdm system of the present invention.

Choosing analogue system is the OFDM Systems that adopts QPSK modulation, and order simulated environment is MATLAB2011b, adopt computer be have Intel Pentium Dual Core processor, in save as 2.96GB, dominant frequency is the computer of 2.16GHz, its operating system is Windows XP SP3, represents the unit (decibel) of signal to noise ratio with dB.The impulsive noise adopting in emulation produces according to Myddelton Class A model (overlapping index gets 0.2, and the power ratio factor gets 0.01).

Fig. 2 has provided in the time that N gets respectively 16,64,128,256,512 and 1024, and signal to noise ratio is-when 5dB, adopt the comparable situation of the average operating time of the CPU that the inventive method and existing impulse noise suppression method based on SBL consume in the time carrying out ofdm system impulse noise mitigation.As can be seen from Figure 2, adopt the average operating time of the CPU that consumes of the inventive method to be significantly less than to adopt the average operating time of the CPU that the existing impulse noise suppression method based on SBL consumes, be that the inventive method has obviously reduced the time that impulse noise mitigation disturbs, practicality is higher.

Fig. 3 has provided in the time of N=256, under different state of signal-to-noise, adopts after the inventive method and the existing impulse noise suppression method based on SBL the symbol error probability situation of change of ofdm system.Setting the OFDM total number of symbols sending is 5000, and signal to noise ratio excursion is-10～30dB.As can be seen from Figure 3, compared with the existing impulse noise suppression method based on SBL, in the time that signal to noise ratio is less than 10dB, the performance of the inventive method is a little less than the existing impulse noise suppression method based on SBL, in the time that signal to noise ratio is greater than 10dB, the performance of the inventive method is obviously better than the existing impulse noise suppression method based on SBL, and adopts the ofdm system of the inventive method to have the performance gain that exceedes 5dB with respect to the ofdm system that does not carry out impulse noise mitigation.

Claims (2)

1. a pulse interference suppression method for ofdm system, is characterized in that comprising the following steps:

1. at transmitting terminal, ofdm signal vector to be sent is designated as to X, supposes to have M known pilot in X, the vector M in X known pilot being formed is designated as X _t; Then X successively carried out inverse discrete Fourier transformer inverse-discrete and adds Cyclic Prefix, obtaining transmission signal vector; Wherein, the dimension of X is N × 1 dimension, N>2,1<M<N, X _tdimension be M × 1 dimension, T represents the indexed set being made up of the index of the known pilot of the M in X, the length of T is M;

2. supposition is known for the state information of the channel of signal transmission, transmitting terminal passes through transmission transmission signal vector to receiving terminal, the impulsive noise vector sum background noise vector correspondence that transmission signal vector is produced in transmitting procedure in channel is designated as e and n, wherein, the dimension of unknown impulsive noise vector e and unknown background noise vector n is N × 1 dimension;

3. at receiving terminal, the received signal vector receiving is successively removed to Cyclic Prefix and discrete Fourier transform (DFT), obtain pending received signal vector, be designated as Y, Y forms after by X superimposed pulse noise vector e and background noise vector n; Then from Y, extract at X _tthe signal vector forming after middle superposeed impulsive noise vector e and background noise vector n, is designated as Y _t; According to the state information of channel, remove Y again _tin with X _tcorresponding known signal vector, is only comprised the known vector Z of unknown impulsive noise vector e and unknown two compositions of background noise vector n; Wherein, the dimension of Y is N × 1 dimension, Y _tdimension be M × 1 dimension, the dimension of Z is M × 1 dimension;

4. the estimated value problem of asking for unknown impulsive noise vector e is converted into the estimated value that solves an e and makes l ₂-l ₁norm type target function the problem of value minimum, then will in be expressed as f (e) with functional form, wherein, symbol " || || ₂" for asking for l ₂norm symbol, symbol for asking for l ₂norm square, symbol " || || ₁" for asking for l ₁norm symbol, F _tthe discrete Fourier transform (DFT) matrix that represents known M × N dimension, τ represents regularization parameter, f () is function representation mode;

5. the estimated value that solves an e by iterative manner makes l ₂-l ₁norm type target function value minimum, detailed process is:

5.-1, the value of setting regularization parameter τ is τ=0.1|| (F _t) ^hz|| _∞, the initial value that makes the estimated value of e is e ₀, e ₀=((F _t) ^hf _t) ^-1(F _t) ^hz, and make k represent iterations, the initial value of k is 1, wherein, symbol " || || _∞" for asking for Infinite Norm symbol, (F _t) ^hfor F _tassociate matrix, ((F _t) ^hf _t) ^-1for (F _t) ^hf _tinverse matrix;

5.-2, the evaluation factor-alpha of the k time iteration of definition _k, ${\mathrm{\&alpha;}}_k=\left\{\begin{array}{cc}1& k=1\\ \frac{{\left(S_k\right)}^H{r}_k}{{\left(S_k\right)}^H{S}_k}& k\&NotEqual;1\end{array}\right.,$ Make again wherein, (S _k) ^hfor S _kassociate matrix, S _k=e _k-e _k-1, e _krepresent the estimated value of the e obtaining after the k time iteration, e _k-1represent the estimated value of the e obtaining after the k-1 time iteration, represent to ask for f (e after the k time iteration _k) first derivative values, represent to ask for f (e after the k-1 time iteration _k-1) first derivative values,

$f\left(e_{k-1}\right)=\frac{1}{2}{\left|\right|Z-F_T{e}_{k-1}\left|\right|}_2^2;$

5.-3, according to α _kand u _k, calculate the estimated value of the e obtaining after the k time iteration, be designated as e _k, wherein, max () is for getting max function, symbol " | | " be the symbol that takes absolute value, represent u _kin i element, 1≤i≤N;

5.-4, judgement whether set up, if set up, execution step 5.-5; Otherwise, make k=k+1, then return to step 5.-2 and continue to carry out, wherein, symbol " || || ₂" for asking for l ₂norm symbol, ε is the minimum normal number of setting, "=" in k=k+1 is assignment;

5.-5, finishing iteration process, obtains the final estimated value of e, is designated as wherein, in "=" be assignment;

6. deduct the final estimated value of e with Y the frequency-domain expression obtaining by discrete Fourier transform (DFT), the received signal vector being inhibited after impulse noise interference.

2. the pulse interference suppression method of a kind of ofdm system according to claim 1, is characterized in that getting ε=10 in described step 5.-4 ^-5.