The interpretation method of Turbo code under a kind of channel for impulse disturbances
Technical field
The present invention relates under a kind of channel for impulse disturbances Turbo code interpretation method.
Background technology
At present, the interchannel noise of most of communication scenes is Gaussian noise, but in some scenes, as electromagnetic radiation is folded
Plus, or multiple instantaneous peak signal collisions, impulsive noise is produced, the statistical property of this impulse noise signal is not total
It is Normal Distribution, but is described with based on the Alpha Stable distritations of broad sense central limit theorem.The more commonly used arteries and veins
Interference channel is rushed for symmetrical Alpha Stable distritations interference channel, affect two major parameters of S α S distributions be characterized factor-alpha and
Scale parameter σ.α determines the hangover degree of distribution, (0,2] between value, especially, when α takes 2, be Gauss distribution, α takes 1
When, it is Cauchy's distribution, the probability density function of S α S distribution variables does not show expression in addition, during α < 2, impulsive noise work(
Rate is infinitely great, and the signal to noise ratio of traditional definition is nonsensical, and α is less, trails more serious, and interference of the impulsive noise to channel is bigger.
The dispersion degree of scale parameter σ description distributions, only under same α, the discussion to σ is just meaningful.Comparatively, feature because
Impacts of the sub- α to channel is bigger.Under S α S impulse disturbances channels, the bit log likelihood ratio of Turbo code input is soft with what channel was exported
No longer it is linear relationship between information.In city or indoor wireless communication, power line communication, Digital Subscriber Line, third and fourth generation shifting
It is likely to the presence of impulse disturbances channel in dynamic communication system.Impulse disturbances cause to decline greatly very much to the performance and throughput of system
Subtract, this needs receiving terminal to take into full account the distribution character of noise, Turbo code decoder input bit likelihood ratio is derived again with letter
Corresponding relation between road output Soft Inform ation, carries out pretreatment to channel output signal, the ability for making full use of channel to decode, more
The loss that supplements punching interference brings, improves systematic function.
The content of the invention
Goal of the invention:The purpose of the present invention is to provide a kind of under impulse disturbances channel for the deficiencies in the prior art
The interpretation method of Turbo code, is processed to channel output Soft Inform ation, and after process, the soft amount of gained likelihood ratio passes to Turbo
Code decoder carries out Max-Log-MAP decodings, on the basis of calculating, storage complexity increase seldom, obtains decoding performance
Larger raising.
Technical scheme:In order to realize goal of the invention, the invention discloses Turbo code under a kind of channel for impulse disturbances
The bit log likelihood ratio that the Soft Inform ation that channel is exported is fitted to Turbo code input is entered again by interpretation method using five segmented fitting methods
Row Turbo code is decoded;Wherein, five segmented fitting methods are divided into five interval, and each areas for the information that is input into channel
Interior information is all fitted with straight line.
Preferably, in order to carry out different disposal for different modulated signals to lift treatment effeciency, five broken line is intended
It is legal to comprise the following steps:
(a1) if the Soft Inform ation of channel output is Turbo code coded sequence Jing BPSK modulated signals and impulse noise signal
Linear superposition, counts f to the Soft Inform ation Y of channel output in [- X, X] intervalNAnd f (Y-1)N(Y+1) centrifugal pump, goes forward side by side
One step calculates discrete LLR (Y);
Wherein, fNThe probability density function of () for impulse noise signal Alpha Stable distritation stochastic variables, LLR (Y) are letter
The relational expression of the bit log likelihood ratio that the Soft Inform ation of road output is input into Turbo code:
(a2) information that channel is input into is divided into five areas by the Soft Inform ation discrete relationship according to the channel output for counting
Between, respectively:(-∞,-X],(-X,Ymin),[Ymin,Ymax],(Ymax,X),[X,+∞);Wherein, YminAnd YmaxRespectively LLR
(Y) take corresponding channel during minimum and maximum value and export Soft Inform ation, the information in each interval is fitted with straight line;
(a3) obtaining five segmented fitting methods to the expression formula that channel exports Soft Inform ation is:
Wherein, the slope of curve of LLR (Y) when P is Y=0, maximums of the H for LLR (Y), the expression formula of a and b are respectively:
The bit log likelihood ratio that Clip (Y) is input into for Turbo code.
Preferably, in order to further lift decoding efficiency, the process of the decoding is comprised the following steps:
(b1) determine whether iteration first, if the prior information of the first component decoder is then given just by iteration first
Value 0, coupling system information and check information enter row decoding, obtain the first outer cover letter breath, and become output after QPP interweaves first
Information is tested, and is sent to second component decoder;
Wherein, the system information is the bit log likelihood ratio in step (a3);
(b2) second component decoder is according to the system information and check information after intertexture and combination output prior information is entered
Row decoding, obtains the second outer cover letter breath, and deinterleaving obtains prior information and sends to the first component decoder;
(b3) judge whether to reach the iterationses of regulation, if reaching, according to the bit of second component decoder output
Log-likelihood ratio sequences are deinterleaved and hard decision, obtain decoding result;Step (b1) is entered otherwise.
Preferably, in order to complete the process work to high order modulation signal, if the useful signal in step (a1) is height
Rank modulated signal, then will send sequence and be divided into I roads and Q roads two-way is transmitted respectively, and be superimposed with impulsive noise respectively, first carry out
Five segmented fitting methods, then carry out Turbo code decoding.
Beneficial effect:The present invention compared with prior art, believe by input of the five segmented fittings method to Turbo code decoder
Breath is processed, and is taken full advantage of the information, is further increased decoding performance.Described Turbo code interpretation method, it is only necessary to deposit
Store up few parameters optimization information and a small amount of multiply-add operation is done to channel output Soft Inform ation, in computation complexity or time complexity only
Under conditions of increasing seldom, decoding performance is significantly improve, so as to improve the throughput of whole system.
Description of the drawings
Fig. 1 is the Turbo code decoder structural scheme of mechanism of the present invention;
Fig. 2 is the corresponding relation between embodiment Turbo code decoder input bit likelihood ratio and channel output Soft Inform ation
Figure;
Using the optimization of five segmented fitting methods and the decoding effect not optimized when Fig. 3 is 1.2 for the channel characteristics factor-alpha of embodiment
Fruit is schemed;
Fig. 4 is using decoding schematic flow sheet of the Turbo interpretation methods of the present invention to qpsk modulation signal.
Specific embodiment
Below in conjunction with the accompanying drawings and embodiment is further described to the present invention.
It is 1024 from the code length in LTE, 1/3 rate codewords of standard, modulation system adopt QPSK, Turbo code decoding side
Method is using simplified log-domain maximum posteriori decoding.The I roads Q roads two paths of signals obtained after modulation is superimposed with into pulse respectively
Noise, then respectively two-way information is processed using five segmented fitting methods, after the two paths of signals after process is deinterleaved as
Turbo code decoder input bit likelihood ratio.Turbo code decoder architecture is as shown in figure 1, YsFor input systematic bits,For the two-way check bit of input, DEC0 and DEC1 is the first component decoder and second component decoder;Decoded
Cheng Wei:An inner iteration is carried out, inner iteration first is determined whether, if inner iteration first, then the prior information to DEC0
Initial value 0 is given, otherwiseValue be that second component decoder DEC1 sendsValue, with reference to YsWithEnter row decoding, obtain
First outer cover letter breathAnd become output prior information after QPP interweavesAnd it is sent to DEC1;After DEC1 is according to intertexture
YsWithAnd combine output prior informationEnter row decoding, obtain the second outer cover letter breathAnd deinterleave intoIt is sent to DEC0,
So far an internal iterative decoding is completed, after internal iterative decoding several times, according to the bit log-likelihood ratio of DEC1 outputs
Sequence is deinterleaved and hard decision, finally gives decoding result.As CRC check position, Ke Yi are carried on each code block
A CRC check is carried out to decoding result at the end of inner iteration every time:If by verification, being effective code word, then terminating in advance
Decoding, output decoding result;If not validated, check whether the inner iteration number of times for reaching regulation, if reaching, then after
Continuous inner iteration next time, otherwise terminates decoding.By this process, it is possible to achieve terminate Turbo decodings in advance, saved system money
Source.
For channel model, as shown in Fig. 2, pulse pulse interference channel different from traditional additive white Gaussian noise channels
Under, no longer it is linear relationship between the bit log likelihood ratio of Turbo code input and the Soft Inform ation of channel output, linearly interval shortens,
Only linear approximate relationship near null value is exported in channel.And channel characteristics factor-alpha is less, linearly interval is less, i.e. impulse disturbances
Interference of the noise to system is stronger, and for same characterization factor α, the scale parameter α of channel is bigger, the discrete journey of impulsive noise
Degree is bigger, and linearly interval is less, and the interference to system is stronger.Using five segmented fitting methods, through numerical statistic, α point is given
Not Wei 1.0, under 1.2,1.4,1.8, the value table of combined parameters (α, σ, P, H, a, b), as shown in table 1.Wherein, when P is Y=0
The slope of curve of LLR (Y), maximums of the H for LLR (Y), the expression formula of a and b is respectively once polynomial in the two intervals
Slope and intercept parameter.
Table 1
alpha |
sigma |
P |
H |
a |
b |
1.2 |
0.64 |
2.828 |
2.51 |
-0.3685 |
2.7484 |
1.4 |
0.64 |
2.922 |
2.673 |
-0.4374 |
3.2266 |
1.6 |
0.70 |
2.350 |
2.777 |
-0.4832 |
3.6602 |
1.8 |
0.80 |
1.832 |
2.938 |
-0.5960 |
4.5595 |
As shown in figure 3, when α is 1.2 without optimization using reception under traditional additive white Gaussian noise noisy communication channel
Turbo code decoding performance under scheme and using the decoding performance simulation result comparison diagram under five segmented fitting schemes, in figure
NoClipping is that unoptimizable decodes effect curve, and 5-line is to decode effect curve using five segmented fitting methods.Due to α<2,
Impulse noise power is infinitely great, and the signal to noise ratio of traditional definition is nonsensical, and the abscissa of simulation result figure is taken as the yardstick of channel
Parameter σ, bit error rate of the vertical coordinate for Turbo code decoder for decoding result.Optimized using five segmented fitting proposed by the present invention
After scheme, decoding performance has and is greatly lifted relative to conventional receiver scheme.