CN101222233A - Turbo coding and decoding method based on unitary space-time symbol - Google Patents

Abstract

The invention provides a Turbo coding/decoding method based on time space symbol and following the symbol level Turbo in the channel of additive white Gaussian noise (AWGN). The method comprises the following steps: grid chart merging is completed within given coherence time; based on the merged grid chart, only the input and output after the merging needs to be taken into consideration during coding and decoding; the bit level processing of convolution code is converted into symbol level processing, thereby converting a binary system into a multiple system; demodulation and decoding are carried out at the same time at a receiving end; decoding of the demodulated soft decision information is completed; finally, the decoded external information is used as the priori information to participate in demodulation once more, thereby forming iteration. The invention adopts the method of combining time space modulation with Turbo code and applies multiple-system Turbo code thought to time space symbol; moreover, the invention adopts iteration MAP to improve coding gain; in addition, iteration decoding ensures that error performance and transmitting power reach an acceptable degree.

Description

Turbo coding and decoding method based on unitary space-time symbol

Technical field

The present invention relates to a kind of coding and decoding method of MIMO (multiple-input and multiple-output) system when being used for not having channel condition information, can be used for realizing under the next generation wireless communication system high speed mobile condition choice criteria of proper communication.

Background technology

Along with the increase of user for two-forty, the requirement of high reliability radio communication under the high-speed mobile condition, traditional empty time-code based on channel estimating often can not reach performance requirement, and the applicable elements that need not the difference empty time-code of channel condition information to be fading coefficients difference between adjacent-symbol can ignore, and there is restriction in the antenna amount that is suitable for.Therefore, press under the rapid fading condition research about the empty time-code that need not channel information.

The calculating of mimo system channel capacity when modulation came from no channel condition information when the chief of a tribe was empty.Marzetta (Marseille moral) and Hochwald people such as (Howards) point out after deliberation: under no channel condition information condition, signal such as should obey at the distribution of side's property during up channel capacity empty.The character of side's property random matrixes such as foundation and chief of a tribe's matrix can get through a series of derivations: Optimal Signals should have following form: S=Φ V=[v ₁φ ₁, v ₂φ ₂..., v _Mφ _M], wherein Φ is chief of a tribe's matrix of T * side's property distributions such as M dimension obedience, and V is a real diagonal matrix, and T and M are respectively channel coherence time and number of transmission antennas.Along with the increase of T, the every value of diagonal matrix V is tending towards

On this basis, the definition of modulation when people such as Hochwald and Marzetta proposes chief of a tribe's sky first: in the quick flat fading environment of Rayleigh, when transmitting-receiving two-end does not all have channel condition information, under the enough big condition of channel T coherence time M or T＞M and receiving terminal signal to noise ratio ρ, optimum transmits and can be expressed as $S_l=\sqrt{T}{\mathrm{\Φ}}_l$ Form.Here, Φ _l∈ { Φ ₁, Φ ₂..., Φ _LAnd L=2 ^RT, wherein L and R are respectively the quantity and the band efficiency of signal in the constellation.Received signal can be expressed as $X=\sqrt{\mathrm{\ρ}/M}\mathrm{SH}+W,$ X is T * N dimension received signal matrix, and H is that M * N ties up channel matrix, and W is that T * N ties up additive noise matrix, and N is a reception antenna quantity.Demodulating algorithm can be expressed as when empty based on the incoherent chief of a tribe of maximum-likelihood criterion:

The structure of constellation also was a problem in the research when chief of a tribe was empty, and a kind of systematized simple structure method is: Φ _l=Θ ^L-1Φ ₁, Φ ₁Be T * M dimension chief of a tribe matrix, can be listed as and constitute by optional M from T * T dimension DFT matrix that Θ is a diagonal matrix, its L power is T * T dimension unit matrix I _TIt must be noted that: the error rate was higher when the chief of a tribe was modulated at low signal-to-noise ratio when empty, and diversity performance is very poor, and this has also verified the precondition that its definition proposes on the other hand.

Classical binary system Turbo code (Parallel Concatenated Convolutional Code) is proposed in 1993 by C.Berrou people such as (cypress sieve), at first the source information bit is carried out the systematic recursive convolutional coding, the source bit carries out once same coding again after interweaving, through multiplexer, each information bit is followed corresponding check bit, forms systematic code.The decoding end adopts iterative decoding, progressively reduces bit error rate.Because of the excellent properties under its low signal-to-noise ratio condition, this coded system is the once change of the communications field.Yet the method is a binary system, and based on AWGN (additive white Gaussian noise) channel, so band efficiency is not high, and mis-behave is serious in fading channel, but in various degree existence decline in the practical application.

The at present existing scheme that modulation combines with Turbo code when empty with the chief of a tribe all is that bit-level is handled, and will add a channel interleaver in addition behind the coding, and receiving terminal is separated to be in harmonious proportion to decipher and carried out respectively, and external information exchanges between decoder and demodulator.Also can reach good performance in theory, but practical application is very difficult.For example, work as T=9, during R=1, constellation is by 2 ⁹Individual chief of a tribe's matrix is formed, and the group of whenever translating a T symbol even do not add coding, still will carry out 2 when adopting maximum-likelihood decoding ⁹Inferior computing.Therefore the amount of calculation of Turbo code own is also very big in addition, and such association schemes are not had a practical significance.All relate to the multiple transmit antennas situation though be fruitful, other condition never is discussed when constant, the performance difference that the number of transmission antennas difference is caused.

Summary of the invention

The present invention is directed to the deficiency of the technical scheme existence that modulation combines with Turbo code when having chief of a tribe's sky now, a kind of Turbo coding and decoding method based on unitary space-time symbol that does not have error performance in the time of can improving the mimo system low signal-to-noise ratio under the channel condition information condition is provided, when other condition homogeneous phase while, under the condition of modulation in conjunction with advanced error correcting code, the transmit diversity effect does not demonstrate fully just as yet can reach preferable performance when the chief of a tribe was empty.

Turbo coding and decoding method based on unitary space-time symbol of the present invention is a coding and decoding method of following symbol stage Turbo code in additive white Gaussian noise (AWGN) channel, in given coherence time, merge grid chart, based on the grid chart that merges, compile, input and output after only considering during decoding to merge, the bit-level processing of convolution code is converted into the symbol level processing, thereby binary system is converted into multi-system, at receiving terminal, separate to be in harmonious proportion to decipher and carry out simultaneously, the soft decision information that demodulates is deciphered, the external information that decoding obtains participates in demodulation again as prior information, forms iteration.

Detailed process of the present invention is as described below:

Follow the coding and decoding method of symbol stage Turbo code in the awgn channel, in given coherence time, merge grid chart.If the shift register quantity of component coder is v, then the number of states in the grid chart is 2 ^vThe selection of component code constraint length should guarantee that parallel route does not appear in grid chart, and promptly along with the increase of T, constraint length also should increase thereupon.After considering that the component coder constraint length is increased to a certain degree, systematic function improves and is not obvious, and the receiving terminal decoding complexity sharply increases, so the present invention is particularly useful for fast fading channel, is not more than 5 with T usually and is advisable.

The composition of the encoder among the present invention as shown in Figure 2, and the main distinction of common Turbo code is the interleaver between the component coder, interleaver among Fig. 2 work in coherence time T be the symbol level of length, before and after promptly interweaving, each is that the group internal bit order of unit is constant with the length T.According to the grid chart that merges, the input of Turbo encoder, output all can be considered the multi-system symbol, and after the chief of a tribe is modulated into the unitary space-time symbol emission when empty.

Receiving terminal decoding adopt be similar to BCJR ((because of L.Bahl (Ba Haer), J.Cocke (examine can), F.Jelinek (Jie Lin can) and J.Raviv proposition such as (upright husband) people and gain the name)) multi-system MAP decoding algorithm.If every group is handled F receiving matrix, purpose is to handle following log-likelihood ratio

$\mathrm{\Λ}\left(S_k^s\right|X_1^F)\∝\ln \left[\underset{(s^{\′},s){\→S}_k^s=S\left(l\right)}{\mathrm{\Σ}}P(s^{\′},s,X_1^F)\right]---\left(1\right)$

Wherein, $P(s^{\&prime;},s,X_1^F)={\mathrm{\&alpha;}}_{k-1}\left(s^{\&prime;}\right)\&CenterDot;{\mathrm{\&gamma;}}_k(s^{\&prime;},s)\&CenterDot;{\mathrm{\&beta;}}_k\left(s\right),$ Three of the equation right-hand member is respectively forward recursion, branch metric and reverse recursion.Represent that as last target " s " this symbol is a system symbol.Forward direction, reverse recursion can be expressed as respectively: ${\mathrm{\&alpha;}}_k\left(s\right)=\underset{\mathrm{all}{s}^{\&prime;}}{\mathrm{\&Sigma;}}{\mathrm{\&gamma;}}_k(s^{\&prime;},s){\&CenterDot;\mathrm{\&alpha;}}_{k-1}\left(s^{\&prime;}\right)$ With ${\mathrm{\&beta;}}_{k-1}\left(s^{\&prime;}\right)=\underset{\mathrm{all\; s}}{\mathrm{\&Sigma;}}{\mathrm{\&beta;}}_k\left(s\right){\&CenterDot;\mathrm{\&gamma;}}_k(s^{\&prime;},s),$ " s " in the bracket represents the number of state indexes in the grid chart at this moment.When each coding anterior displacement register was all done " zero clearing " processing, the initial value of forward recursion was always $\left\{\begin{array}{c}{\mathrm{\&alpha;}}_0({\mathrm{<figure-callout\; id="0"\; label="S"\; filenames="S2007101158652E00031.png"\; state="\{\{state\}\}">S</figure-callout>}}_0=0)=1\\ {\mathrm{\&alpha;}}_0({\mathrm{<figure-callout\; id="0"\; label="S"\; filenames="S2007101158652E00031.png"\; state="\{\{state\}\}">S</figure-callout>}}_0\&NotEqual;0)=0\end{array}\right.;$ When " making zero " processing is done in the coding ending, $\left\{\begin{array}{c}{\mathrm{\&beta;}}_F(S_F=0)=1\\ {\mathrm{\&beta;}}_F(S_F\&NotEqual;0)=0\end{array}\right.,$ Otherwise the initial reverse recursion value of each state equates.Branch metric can be calculated as γ _k(s ', s)=∑ P (s|s ') P (S _k| s ', s) P (X _k| S _k).Concrete representation class is similar to general T urbo sign indicating number, only expresses the 3rd here, and all the other repeat no more.Because the decline that each unitary space-time symbol experienced is separate, then the 3rd of following formula can be expressed as $P\left(X_k\right|S_k)=P\left(X_k^s\right|S_k^s)\&CenterDot;P\left(X_k^p\right|S_k^p).$ Because the receiving symbol conditional probability when not having channel condition information can be expressed as:

So:

A _kBe the public keys of each possible path of synchronization, can offset.Therefore

$\mathrm{\&Lambda;}\left(S_k^s\right|X_1^F)$

$={\mathrm{\&Lambda;}}^a\left(S_k^s\right)+{\mathrm{\&Lambda;}}^c\left(S_k^s\right)+{\mathrm{\&Lambda;}}^e\left(S_k^s\right)$ (4)

Here, Λ ^a, Λ ^cAnd Λ ^eIt is respectively the external information of prior information, channel metrics and decoding gained.For the first time during iteration, prior information can all be made as zero, deciphers the prior information of the external information of gained as next iteration decoding at every turn.

The method that modulation combined with Turbo code when the present invention adopted the chief of a tribe empty is applied to unitary space-time symbol with multi-system Turbo code thought, and with iterative MAP (maximum a posteriori probability).Though coding is that bit-level is carried out, from the angle that grid merges, also can be considered is the unitary space-time symbol level.The soft information that demodulator passes to decoder is based on the correlation between unitary space-time symbol.Employing is similar to the algorithm of BCJR, has improved coding gain.Adopt iterative decoding, can make error performance and transmitting power reach practical acceptable degree.Error performance when the present invention can improve the mimo system low signal-to-noise ratio under no channel condition information condition.

Description of drawings

Fig. 1 is the system principle diagram that the present invention is based on the Turbo coding and decoding method of unitary space-time symbol.

Fig. 2 is the composition frame chart of encoder among the present invention.

Fig. 3 is an iterative decoder composition frame chart among the present invention.

Fig. 4 is under single transmit, single reception antenna condition, the comparison diagram of bit error rate curve under the various code checks.

Fig. 5 is under single transmit, single reception antenna condition, the comparison diagram of iteration effect.

Fig. 6 is a receive diversity errored bit performance map.

Fig. 7 is a transmit diversity errored bit performance map.

Embodiment

As shown in Figure 1, at first every T source signal is formed a frame, so that the unitary space-time symbol level is handled.Based on the grid chart that merges, input and output after compiling, only considering to merge during decoding are handled the bit-level of convolution code being converted into symbol level and handling, thereby binary system are converted into multi-system, coding back symbol carries out modulating when the chief of a tribe is empty, via M transmission antennas transmit.At receiving terminal, demodulator and decoder are done as a whole, carry out simultaneously, and the soft decision information that demodulates is sent into decoder, and the external information that decoding obtains is sent into demodulator again as prior information and participated in demodulation, forms iteration.Signal after the decoding is handled through removing frame, becomes baseband signal.

Among the present invention based on the concrete composition of the encoder of unitary space-time symbol as shown in Figure 2.Each component coder among Fig. 2 all is the systematic recursive convolutional encoder, and source signal is sent into second component coder after interweaving through symbol level.Can control code check by deleting surplus device deletion school inspection symbol.The effect of multiplexer is to follow corresponding school inspection symbol closely after making each system symbol, to form systematic code, S among the figure _k ^sRepresent k system symbol, S _k ^pRepresent k the checking symbol that encoded back forms.The effect of deleting surplus device is to improve code check, promptly improves band efficiency.System symbol and checking symbol are through forming systematic code S behind the multiplexer _k, promptly each system symbol heel is with corresponding checking symbol.Different with classical Turbo code, the processing here no longer is based on bit-level, and is based on symbol level.Among the present invention based on the concrete composition of the iterative decoder of unitary space-time symbol as shown in Figure 3, among Fig. 3, X _k ^sAnd X _k ^{S '}Represent k the system symbol that receiver receives respectively and interweave after k system symbol.X _1k ^pAnd X _2k ^pBe respectively corresponding to k school of the 1st component coder and examine symbol and examine symbol corresponding to k school of the 2nd component coder.L _e(S _k) external information of k symbol of expression after the component decoder for decoding, Λ (S _k) be through all deciphering the final log-likelihood ratios of k the symbol in back.During decoding, to each the frame integrated treatment that receives, its k system symbol and first group of school inspection symbol are sent into first component decoder simultaneously, because be decoding for the first time at this moment, so impartial general appearance of all symbols of hypothesis, adopt foregoing MAP decoding algorithm, obtain the external information of each symbol.System symbol is sent into second component decoder simultaneously with second group of school inspection symbol after interweaving.Also send into second component decoder by the external information that decoding for the first time obtains through interweaving, as its prior information fellowship decoding.Like this, finish decoding for the first time.Can send into first component decoder by the external information that second component decoder obtains through interweaving, form iteration.By certain criterion or preset value, can determine iterations.At last, all the log-likelihood ratio that obtains of decoding through deinterleaving again hard decision promptly obtain baseband signal.

Emulation experiment is verified the performance of the inventive method by following several respects: coding validity, iteration effect and diversity.Wherein, comparatively special because of transmit diversity situation, carry out at last.In all experiments, component code is the systematic recursive convolutional sign indicating number, and as there not being explanation, encoder matrix is (7,5) ₈, for simplicity, below performance during main examination T=2.

1) coding validity

Under single transmit, single reception antenna condition, the curve of bit error rate as shown in Figure 4 under the various code checks.

Modulation can be handled the fastest situation of rate of fading when T=2 was chief of a tribe's sky, and as can be seen from Figure 4, for reaching acceptable errored bit performance, the modulation required signal-to-noise ratio was high, impracticable when the simple chief of a tribe was empty.When adopting code check to be 1/3 can be 10 in bit error rate then ^-5The time obtain the gain of about 35dB; When employing is deleted surplus back code check and reached 1/2, still can obtain the gain about 25dB.

2) iteration effect

Under single transmit, single reception antenna condition, the iteration effect as shown in Figure 5.

When code check is 1/3, adopt iterative decoding, the errored bit performance can further improve.No matter be bit error rate or required signal-to-noise ratio, all can be practical application and accept.Iterative process and classical Turbo code are similar: along with the increase of iterations, the improvement of performance is fewer and feweri, and bit error rate tends towards stability.Experiment shows that when adopting the MAP algorithm of no approximate processing, iterations is very little greater than 4 times performance raising, can ignore.

3) receive diversity

Receive diversity errored bit performance contrasts with Fig. 5 as shown in Figure 6, and multiple receive antenna can improve systematic function once more as can be seen.This moment, reception antenna quantity was N=2, and code check still is 1/3.In the modulation, the quantity of reception antenna was not influenced by channel T coherence time when the chief of a tribe was empty, so use comparatively flexible; From the derivation of front as can be seen, it is big unlike increasing the required amount of calculation of equal number transmitting antenna to increase reception antenna.Therefore, receive diversity is a kind of comparatively economic method.

4) transmit diversity

When physical condition limited the use of receive diversity, transmit diversity was used in essential consideration.Because number of transmission antennas should not surpass half of coherence time of channel, so the employing multiple transmit antennas requires T=4 at least.

As definition required, the chief of a tribe is modulated under the high s/n ratio condition when empty just can embody emission diversity performance preferably.When the simple chief of a tribe is modulated at low signal-to-noise ratio when empty, adopt multiple transmit antennas can improve bit error rate on the contrary.The most outstanding advantage of Turbo code can reach good errored bit performance when being low signal-to-noise ratio, so in the associated methods that proposes, transmit diversity does not embody its good effect as yet and just reached preferable performance.As can be seen from Figure 7, compare with single transmit antenna situation, although after the diversity during low signal-to-noise ratio systematic function poor all the better, the amplitude that its performance improves with signal to noise ratio is bigger during than no diversity.That is to say that if the errored bit performance is had requirements at the higher level, along with the increase of signal to noise ratio, emission diversity gain will manifest gradually.

Claims (1)

1. Turbo coding and decoding method based on unitary space-time symbol, it is characterized in that: be the coding and decoding method of following symbol stage Turbo code in the additive white Gaussian noise channel, in given coherence time, merge grid chart, based on the grid chart that merges, compile, input and output after only considering during decoding to merge, the bit-level processing of convolution code is converted into the symbol level processing, thereby binary system is converted into multi-system, at receiving terminal, separate to be in harmonious proportion to decipher and carry out simultaneously, the soft decision information that demodulates is deciphered, and the external information that decoding obtains participates in demodulation again as prior information, forms iteration.

Images