CN106254292A - Reduction PAPR algorithm based on the distribution of stochastic filtering device in GFDM system - Google Patents

Abstract

The present invention is claimed in a kind of GFDM system reduction PAPR algorithm based on the distribution of stochastic filtering device, relates to wireless communication system.The same with all of multicarrier system, due to output signal superposition, GFDM system still has high PAPR.Although the method reducing PAPR in ofdm system has a lot, but, due to GFDM system, directly use in ofdm system that to reduce the method effect of PAPR unsatisfactory.For the problem that research exists at present, the construction features of wave filter can be selected flexibly according to GFDM system, a kind of method proposing new reduction system PAPR, by the wave filter different to every subcarriers random assortment, phase equalization when being intended to destroy Signal averaging thus effectively suppress PAPR, and the distortion of signal will not be caused.Theory analysis and simulation result confirm the superiority of this algorithm.

Description

Reduction PAPR algorithm based on the distribution of stochastic filtering device in GFDM system

Technical field

The present invention relates to wireless communication field, particularly relate to the skill of reduction PAR peak to average ratio in filter bank multi-carrier technology Art.

Background technology

At present, the standard of a lot of radio communications all relies on OFDM, and it is as a kind of important multi-transceiver technology, will be at a high speed Data stream is divided into several parallel low rate data streams of transmission on different subcarriers, is effective against intersymbol interference (ISI), but OFDM still has limitation, to this end, the novel multi-carrier modulation technology such as GFDM arises at the historic moment.GFDM is as 5G Alternative waveform, there is the advantage that a lot of OFDM does not has: data symbol can be by across time domain and the two-dimentional modular structure of frequency domain Propagate, employ adjustable pulse shaping filter and make system have the highest motility etc..

But the same with OFDM multicarrier system, owing to transmitting terminal output signal is overlapped mutually, may at a time produce Raw bigger peak power, thus GFDM system can bring higher papr, is called for short peak-to-average force ratio (PAPR).Theoretical The quantity of upper minimizing carrier wave can largely reduce the PAPR of system, and the most infeasible, so Under conditions of multicarrier, the PAPR of high degree reduction system is only the key of solution problem.Although ofdm system reduces PAPR Method have a lot, but, due to the construction features of GFDM system, directly use in ofdm system the method effect reducing PAPR Unsatisfactory.

Document [Sendrei L, S,Michailow N,et al.Iterative receiver for clipped GFDM signals[C]//IEEE 24th International Conference Radioelektronika.2014, pp.1-4.] utilize iterative receiver that GFDM reception signal is processed (Clipped- GFDM), this can cause signal distortion, although successive ignition can effectively eliminate the nonlinear noise caused due to amplitude limit, but Iterations is the most, and operand is the most complicated；Document [Sharifian, Z, Omidi, M.J, Farhang, A, et al.Polynomial-based compressing and iterative expanding for PAPR reduction in GFDM [C] //IEEE 23rd Conference on Electrical Engineering, 2015, pp.518-523.] be given It is a kind of that based on multinomial companding algorithm (PCT-GFDM), it is to map one to one, and at receiving terminal by iterative receiver mode Recover transmitting terminal signal, but this algorithm also can increase operand while reducing PAPR and sacrifice BER performance.

GFDM multicarrier system model is as shown in Figure 1.Binary sequence b completes corresponding constellation mapping by QAM modulation, Obtain sequence d for complex data symbol composition, after also (S/P) converts and up-sample (Upsampling) through string, form K road bag Parallel data stream d containing M symbol_k={ d_0,k,d_1,k……d_M-1,k}^T, (k=0,1 ...., K-1), every road signal is with each The Pulse shaped filter of self-loopaBeing filtered, filter period is MN, then by the mid frequency of respective subcarrier Modulate superposition again to obtain sending signal x [n], be eventually adding Cyclic Prefix, send into signal.

$x\[n\]=\underset{k=0}{\overset{K-1}{\mathrm{\Σ}}}{x}_k\[n\]=\underset{k=0}{\overset{K-1}{\mathrm{\Σ}}}\underset{m=0}{\overset{M-1}{\mathrm{\Σ}}}{d}_{m,k}\mathrm{\δ}(n-mN)\⊗\tilde{g}\[n\]e^{j2\mathrm{\π}\frac{kn}{N}}---\left(1\right)$

Wherein,Represent the cyclic convolution about n, n ∈ [0, MN-1], d_m,kRepresent the m-th of transmission on kth carrier wave Complex data symbol.

According to document [Michailow N, Gaspar I, Krone S, et al.Generalized frequency division multiplexing:Analysis of an alternative multi-carrier technique for next generation cellular systems[C]//International Symposium on Wireless Communication Systems.IEEE, 2012:171-175.] GFDM module can realize with IFFT/FFT, time domain grasped Be converted to frequency-domain operations, as in figure 2 it is shown, then formula (1) can be expressed as:

$x=W_{MN}^H\underset{k}{\mathrm{\Σ}}{P}^{\left(k\right)}{\mathrm{\ΓR}}^{\left(L\right)}{W}_M{d}_k---\left(2\right)$

Wherein, W_Md_kRepresent the FFT of M point；In a frequency domain, up-sampling processes and shows as signal replication, replicates matrix R^(L)It is By L I_MIt is in series, i.e. R^(L)={ I_M,I_M,......,I_M}^T, I_MUnit matrix for M × M；Electric-wave filter matrix Γ=diag (W_LMG), g is the time sampling vector of filter pulse；Frequency conversion is that signal is multiplied by a permutation matrix P at frequency domain representation^(k), P^(k)According to P⁽¹⁾={ I_LM,0_LM,0_LM,...}^T, P⁽²⁾={ 0_LM,I_LM,0_LM,...}^TRule analogize, wherein 0_LMFor LM × The null matrix of LM.

Further, send signal to be expressed as with linear matrix:

X=Ad (3)

Here, A represents the modulation matrix of a MN × KM.

After removing Cyclic Prefix (CP), the vector being made up of reception sample of signal can be write as:

Y=Hx+n (4)

Wherein, H is the impulse response function of channel,Represent that average is 0 and variance isAdditivity high This white noise vector ,～represent equivalent；

At receiving terminal, through frequency domain equalization (FDE) channel distortion compensated and obtainIn order to eliminate or suppress due to In the nonopiate inter-carrier interference (ICI) caused of intercarrier, generally utilize matched-filter receiver (MF), zero-forcing receiver (ZF), three kinds of linear GFDM receivers such as least mean-square error receiver (MMSE) rebuild GFDM systems and send data, use I_MNIt is The unit matrix of MN × MN, has the most respectively:

${\hat{d}}_{MF}=A^H\tilde{y}---\left(5\right)$

${\hat{d}}_{ZF}={\left(A^{H}A\right)}^{-1}{A}^H\tilde{y}---\left(6\right)$

${\hat{d}}_{MMSE}={(A^{H}A+{\mathrm{\σ}}_n^2{I}_{MN})}^{-1}{A}^H\tilde{y}---\left(7\right)$

The PAPR of GFDM system is represented by:

$PAPR=\frac{P_{peak}}{P_{average}}=10{\log }_{10}\frac{\underset{n\∈\[0,MN-1\]}{max}\left\{{\left|x\[n\]\right|}^2\right\}}{E\left\{{\left|x\[n\]\right|}^2\right\}}---\left(8\right)$

Wherein E{ } represent computing of averaging.For convenience, permissible according to the signal after being modulated on formula (1) kth carrier wave It is expressed as:

$x_k\[n\]=\underset{m=0}{\overset{M-1}{\mathrm{\Σ}}}{d}_{m,k}\tilde{g}(n-mN)e^{j2\mathrm{\π}\frac{kn}{N}}---\left(9\right)$

Owing to intersymbol is separate, so x_kBetween [n] the most separate, thus have:

E{x_k[n] }=0

$D\{x_k\[n\]\}=E\{x_k\[n\]x_k^{*}\[n\]\}={\mathrm{\σ}}_d^2\underset{m=0}{\overset{M-1}{\mathrm{\Σ}}}{\tilde{g}}^2(n-mN)={\mathrm{\σ}}^2---\left(10\right)$

X as can be seen from the above equation_kAverage and the variance of [n] are unrelated with k, according to central limit theorem, when K is enough Time big, x [n] obedience average is 0 and variance is 2 σ_n ²=K σ²Multiple Gauss distribution, wherein σ_n ²Represent x [n] real part and the side of imaginary part Difference.

The performance of system PAPR can with compensating error cumulative function (CCDF) represent, it can calculate PAPR exceed to The probability of threshold value γ gone out.By analysis above, the envelope of x [n] | x [n] | Rayleigh distributed, power | x [n] |²Obey Degree of freedom is the χ of 2²Distribution.| the x [n] | that makes S=², then the probability density function of S is:

$f_S\left(s\right)=\frac{1}{{\mathrm{K\σ}}^2}{e}^{-\frac{s}{{\mathrm{K\σ}}^2}},s>0---\left(11\right)$

OrderThen the probability density function of T is:

f_T(t)=K σ²f_S(Kσ²T)=e^-t, t ＞ 0 (12)

Then have

$P(t<\mathrm{\&gamma;})={\&Integral;}_0^{\mathrm{\&gamma;}}{f}_T\left(t\right)dt=1-e^{-\mathrm{\&gamma;}}---\left(13\right)$

Owing to each T is independent, so the cumulative function of PAPR (CDF) can be expressed as:

$P(PAPR<\mathrm{\&gamma;})=P\left(\underset{n=0}{\overset{MN-1}{\&cap;}}\right(\frac{{\left|x\&lsqb;n\&rsqb;\right|}^2}{E\left\{{\left|x\&lsqb;n\&rsqb;\right|}^2\right\}}<\mathrm{\&gamma;}\left)\right)={(1-e^{-\mathrm{\&gamma;}})}^{MN}---\left(14\right)$

Therefore, the complimentary cumulative function (CCDF) of the PAPR of GFDM system is:

P (PAPR ＞ γ)=1-(1-e^-γ)^MN (15)

The construction features of wave filter can be selected flexibly to have produced the reason of high PAPR, invention one in conjunction with GFDM system The new algorithm reducing system PAPR based on random assortment wave filter.

Summary of the invention

Present invention seek to address that above problem of the prior art.Propose a kind of method.Technical scheme is as follows:

Reduction PAPR algorithm based on the distribution of stochastic filtering device in a kind of GFDM system, it comprises the following steps:

Step 1, GFDM system initialization, the stochastic generation one sequence C={ c containing K element₀, c₁......c_k,......c_K-1, wherein c_k∈{0,1}；Wherein the size of K is the sub-carrier number size of system；

Step 2, the K subcarriers signal of acquisition GFDM system, select wave filter to be filtered K subcarriers signal also Numbering, works as c_kWhen=0, indicate that K subcarriers signal is filtered by selection the first wave filter；Work as c_kWhen=1, select the K subcarriers signal is filtered by two wave filter, and wherein the first wave filter and the second filter type are different；

Step 3, obtain GFDM signal to through step 2 filtered K subcarriers Signal averaging, calculate GFDM signal PAPR also stores.

Further, described step 2 works as c_kWhen=0, raised cosine filter RC selected by the first wave filter；Work as c_kWhen=1, Root raised cosine filter RRC selected by second wave filter.

Further, the K subcarriers signal packet of GFDM system described in described step 2 parallel data stream containing M symbol d_k={ d_0,k,d_1,k……d_M-1,k}^T, (k=0,1 ...., K-1).

Further, particularly as follows: first step 3 calculates the step of the PAPR of GFDM signal, calculates this N number of data symbols respectively Number power, secondly select maximum power therein, calculate the mean power of GFDM signal, finally, by maximum power The PAPR i.e. obtaining GFDM is compared with mean power.

Advantages of the present invention and having the beneficial effect that:

The basic reason causing PAPR in multi-carrier systems is the continuous superposition of sub-carrier signal, when sub-carrier phase is consistent Time arise that higher peak power.In traditional GFDM system, signal is entered by every subcarriers with identical wave filter Row processes, and this can increase the probability that sub-carrier phase is consistent.To this end, herein according to the phase-frequency characteristic of different wave filter different this Feature, proposes a kind of new method reducing PAPR.By the wave filter different to each subcarrier random assortment, between subcarrier Complex data symbol introduce out of phase offset vector, it is consistent that high degree destroys sub-carrier phase, thus effectively suppresses PAPR, The program is referred to as based on stochastic filtering device allocation algorithm (RAF-GFDM).Select two kinds of common and effective raised cosine filters herein Ripple device (RC) and root raised cosine filter (RRC), so can share with other modules when system realizes.

Compared by the complexity of several different fall PAPR algorithms, it can be seen that margining amplitude technique does not increase extra complexity Degree, but reduce PAPR with the bit error rate performance of sacrificial system；The computation complexity of PCT algorithm is increased slightly, but it is with limit Width method is the same, can increase the bit error rate of system.SLM algorithm is a linear process, will not bring signal distortion, and complexity is also Moderate, but the effect reducing PAPR does not highlight；And algorithm in this paper is linear process equally, though reducing PAPR's Do not have margining amplitude technique and PCT algorithm to highlight in effect, but victory do not increasing extra calculating cost, and to the bit error rate almost without Impact, has higher actual application value.

Accompanying drawing explanation

Fig. 1 is that the present invention provides preferred embodiment GFDM transmitting terminal system model.

Fig. 2 GFDM signal sending end frequency domain processes model

Fig. 3 RAF-GFDM process chart

PAPR performance under Fig. 4 difference rolloff-factor

Fig. 5 PAPR reducing effect comparison diagram

BER Performance comparision under MF in Fig. 6 Gaussian channel

BER Performance comparision under ZF in Fig. 7 Gaussian channel

BER Performance comparision under MMSE in Fig. 8 Gaussian channel

BER Performance comparision under MF in Fig. 9 walking multipath channel

BER Performance comparision under ZF in Figure 10 walking multipath channel

BER Performance comparision under MMSE in Figure 11 walking multipath channel

The BER performance of algorithms of different in Figure 12 Gaussian channel

The BER performance of algorithms of different in Figure 13 walking multipath channel

Table 1 transmitting terminal computational complexity compares

Table 2 simulation parameter

Detailed description of the invention

Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, detailed Carefully describe.Described embodiment is only a part of embodiment of the present invention.

Technical scheme is as follows:

Select two kinds of common and effective raised cosine filters (RC) and root raised cosine filter (RRC) herein, so Can share with other modules when system realizes.Specifically comprise the following steps that

(1) first, stochastic generation one comprises the binary sequence C of K element_k∈ { 0,1}, k ∈ [0, K-1].

(2) then, wave filter is numbered, works as C_kWhen=1, select RCC wave filter that kth subcarriers signal is carried out Filtering, otherwise selects RC wave filter to be filtered kth subcarriers signal.

(3) last, K road-load ripple superposition obtains GFDM signal, calculates PAPR.

Next being analyzed the complexity of RAF-GFDM above, for comparing, we give also OFDM and tradition The complexity of GFDM system, Clipped-GFDM system, PCT-GFDM, SLM-GFDM system etc..With reference to formula (2), generate GFDM The step of signal demand is as follows:

(1) FFT of K M point；

(2) K filtering, once filtering carries out LM CM computing；

(3) plural number of K MN point adds computing；

(4) IFFT of K MN point.

We use CM computing C^MThe complexity of measurement system, so the complexity of traditional GFDM system can represent For:

$C_{GFDM}^M=KM{\log }_{2}M+KLM+MN{\log }_{2}MN---\left(16\right)$

Algorithm based on random assortment wave filter in this paper, compared to traditional GFDM system, simply at signal It is varied from Filtering Processing step, but the computing cost of not increase system, so:

$C_{RAF-GFDM}^M=C_{GFDM}^M=KM{\log }_{2}M+KLM+MN{\log }_{2}MN---\left(17\right)$

As shown in table 2, the complexity giving several different fall PAPR algorithm compares, it can be seen that margining amplitude technique does not increase Add extra complexity, but reduce PAPR with the bit error rate performance of sacrificial system；The computation complexity of PCT algorithm slightly increases Add, but it is the same with margining amplitude technique, the bit error rate of system can be increased.SLM algorithm is a linear process, and signal will not be brought abnormal Becoming, complexity is the most moderate, but the effect reducing PAPR does not highlight；And algorithm in this paper is linear process equally, though The effect reducing PAPR does not has margining amplitude technique and PCT algorithm highlight, but victory is not increasing extra calculating cost, and right The bit error rate has little to no effect, and has higher actual application value.

Herein by wave filter RC and RRC different to every subcarriers random assortment, it is intended to destroy K subcarriers signal The concordance of phase place, makes the probability occurring peak power when of the continuous superposition of each sub-carrier signal reduce.In order to analyze herein The performance of method, iteration margining amplitude technique (Clipped-GFDM) that we propose with tradition GFDM, document [15], multinomial companding method (PCT-GFDM) carry out emulation compare with Choose for user algorithm (SLM-GFDM).Channel includes Gaussian channel and walking multipath letter Road, is respectively adopted tri-kinds of reception modes of MF, ZF, MMSE, and other simulation parameters herein are as shown in table 2.

Fig. 3 is the impact on RAF-GFDM system PAPR of the rolloff-factor of wave filter.It was noticed that RAF-GFDM system PAPR performance and the rolloff-factor (roll-off) of institute's selecting filter relevant, be not to be the bigger the better, be not the smaller the better, But general impacts are not the biggest.So we need the RAF-selecting suitable rolloff-factor to make system in actual applications The PAPR performance of GFDM system reaches optimum.

The method of different reduction PAPR is compared by Fig. 4, it was noted that the PAPR of GFDM system is slightly above OFDM, this is owing to GFDM has used more flexible wave filter, i.e. RRC, RC etc., accelerates the attenuation outside a channel of system.Assume to use Be normalized wave filter, then the mean power of GFDM signal will be less than ofdm signal, so the PAPR of GFDM compares OFDM can increased.We are it can also be seen that compare with tradition GFDM simultaneously, and method in this paper can effectively reduce The PAPR of GFDM system.Although it addition, the performance that Clipped-GFDM scheme and PCT-GFDM scheme reduce PAPR is better than herein The algorithm proposed, but both algorithms broadly fall into nonlinear operation, and not only complexity is higher, also can have a strong impact on system BER performance.Further, we also make comparisons with the Choose for user algorithm (SLM) being all linear operation, the PAPR fall of context of methods Low effect is better than SLM algorithm.

Fig. 6 and Fig. 7, respectively research Gaussian channel and walking multipath channel in RAF-GFDM system in this paper MF, ZF, MMSE tri-kinds receive mode BER performance, and with only use a kind of wave filter traditional GFDM system BER performance make compare Relatively.It was noticed that the BER of the no matter RAF-GFDM system under Gaussian channel or walking multipath channel, MF reception mode Performance curve is all between the BER performance curve of traditional GFDM system being used alone RC or RRC.And it should be noted that According to ZF and MMSE receive mode, when signal to noise ratio (SNR) is relatively low, the BER performance curve of RAF-GFDM system still between It is used alone between the BER performance curve of traditional GFDM system of RC or RRC；But the RAF-GFDM system when middle high s/n ratio BER performance can the most nearly, this explanation when signal to noise ratio is higher, BER is bigger by the filters affect of poor-performing.Combine Upper described, algorithm based on random assortment wave filter in this paper is less to the BER performance impact of GFDM system.

Fig. 8 with Fig. 9 is respectively Gaussian channel and uses different reduction PAPR methods in GFDM system in walking multipath channel BER Performance comparision.It can be seen that the Clipped-GFDM method that proposes of document [15] and document [16] propose The BER performance of PCT algorithm has substantially variation, because both algorithms broadly fall into non-linear process, the PAPR reducing system is With sacrifice BER performance as cost；RAF-GFDM algorithm in this paper but can reduce system on the premise of ensureing BER performance The PAPR of system.

Table 1 is that transmitting terminal computational complexity compares

Table 1

Table 2 is simulation parameter

Table 2

The above embodiment is interpreted as being merely to illustrate the present invention rather than limiting the scope of the invention.? After the content of the record having read the present invention, the present invention can be made various changes or modifications by technical staff, and these equivalences become Change and modify and fall into the scope of the claims in the present invention equally.

Claims (4)

1. reduction PAPR algorithm based on the distribution of stochastic filtering device in a GFDM system, it is characterised in that comprise the following steps:

Step 1, GFDM system initialization, the stochastic generation one sequence C={ c containing K element₀,c₁......c_k, ......c_K-1, wherein c_k∈{0,1}；Wherein the size of K is the sub-carrier number size of system；

Step 2, the K subcarriers signal of acquisition GFDM system, select wave filter that K subcarriers signal is filtered and is compiled Number, work as c_kWhen=0, indicate that K subcarriers signal is filtered by selection the first wave filter；Work as c_kWhen=1, select second K subcarriers signal is filtered by wave filter, and wherein the first wave filter and the second filter type are different；

Step 3, obtain GFDM signal to through step 2 filtered K subcarriers Signal averaging, calculate the PAPR of GFDM signal And store.

Reduction PAPR algorithm based on the distribution of stochastic filtering device in GFDM system the most according to claim 1, its feature exists In, described step 2 works as c_kWhen=0, raised cosine filter RC selected by the first wave filter；Work as c_kWhen=1, the second wave filter is selected Root raised cosine filter RRC.

Reduction PAPR algorithm based on the distribution of stochastic filtering device, its feature in GFDM system the most according to claim 1 and 2 It is, the K subcarriers signal packet of GFDM system described in the described step 2 parallel data stream d containing M symbol_k={ d_0,k, d_1,k……d_M-1,k}^T, (k=0,1 ...., K-1).

Reduction PAPR algorithm based on the distribution of stochastic filtering device, its feature in GFDM system the most according to claim 1 and 2 Being that a GFDM signal packet contains N number of data symbol, particularly as follows: first step 3 calculates the step of the PAPR of GFDM signal, respectively Calculate the power of this N number of data symbol, select maximum power therein, secondly, calculate the mean power of GFDM signal, After, maximum power is compared with mean power the PAPR i.e. obtaining GFDM.