KR101076228B1 - Apparauts and method for list sphere decoding in mobile communication system - Google Patents

Abstract

The present invention relates to a mobile communication system, and more particularly, to a method and apparatus for List Sphere Decoding (LSD) in which the complexity of a receiver is reduced in a mobile communication system.
In addition, in the mobile communication system according to the present invention, a method for decoding a list sphere (LSD) of a receiver is calculated based on a cost function of a set of symbol vectors input from a transmitter, and then calculates a margin ratio of a first point and an initial search radius. And setting the initial search radius in consideration of the calculated margin ratio, and performing LSD on symbols included in the set initial search radius.

Description

Apparatus and METHOD FOR LIST SPHERE DECODING IN MOBILE COMMUNICATION SYSTEM}

The present invention relates to a mobile communication system, and more particularly, to an apparatus and a method for List Sphere Decoding (LSD) in which a complexity of a receiver is reduced in a mobile communication system.

MIMO communication system using Multiple Input Multiple Output (MIMO) can increase communication capacity and receiver performance by using multiple transmit antennas and multiple receive antennas to transmit data. However, in the MIMO communication system, since interference exists between layers and random fading of the channel matrix, a problem may occur when detecting a signal.

In particular, when implementing soft decoding in the MIMO communication system, the characteristics of the channel matrix when a spatial multiplexing (SM) method, a zero-forcing (ZF) method, or a minimum mean square error bit (MMSE) method are used This can degrade performance. On the contrary, when the lattice code detection method is used, better performance can be obtained, but the complexity increases exponentially with an increase in the modulation level or the number of transmitting antennas.

In order to improve the above problems, new signal detection techniques have been proposed in MIMO communication systems. A widely used method is List Sphere Decoding (LSD). The LSD scheme lowers the complexity of the conventional grid code detection scheme and has performance according to the size of the list. However, the complexity of the LSD scheme can also be improved in that the size of the list is always fixed. That is, the complexity of the LSD scheme is determined according to the size of a list, an initial search radium for searching for a symbol, and an initial search point. This indicates that the retrieval process of symbols is very diverse and it is difficult to implement a practical hardware design. Accordingly, other types of LSDs are required to make the symbol retrieval process constant and to facilitate hardware design.

Accordingly, there is a need for an optimal LSD scheme that can reduce the complexity of the receiver while approaching the performance of the existing LSD in a MIMO communication system.

The present invention provides an apparatus and method for optimal LSD that can be used in signal detection in a MIMO communication system.

The present invention also provides an apparatus and method for LSD having high performance and low complexity in a MIMO communication system.

The apparatus for List Sphere Decoding (LSD) of a receiver in a mobile communication system according to the present invention includes an LLR calculator for calculating a log-likelihood ratio (LLR) for a set of symbol vectors received from a transmitter, and the cost of the set of symbol vectors. A margin ratio of the first point and the initial search radius is calculated based on a function, the initial search radius is set using the calculated margin ratio, and then LSD is performed on symbols included in the set initial search radius. It includes a decoder.

In addition, in the mobile communication system according to the present invention, a method for LS Sphere decoding (LSD) of a receiver calculates a margin ratio of a first point and an initial search radius based on a cost function of a set of symbol vectors input from a transmitter. And setting the initial search radius by using the calculated margin ratio and performing LSD on symbols included in the set initial search radius.

The present invention can reduce the complexity of the receiver while approaching the performance of the existing LSD by decoding using the LSD in the receiver of the MIMO communication system.

On the other hand various other effects will be disclosed directly or implicitly in the detailed description of the embodiments of the present invention to be described later.

1 is a view showing a mobile communication system to which the present invention is applied;
2 is a view showing an example of a transmitter 101 in a mobile communication system to which the present invention is applied;
3 is a view showing an example of a receiver 226 in a mobile communication system to which the present invention is applied;
4 is a diagram illustrating a process for performing LSD in a detector 237 of a receiver according to an embodiment of the present invention;
5 is a diagram showing the performance and complexity in a mobile communication system according to an embodiment of the present invention.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. It should be noted that in the following description, only parts necessary for understanding the operation according to the present invention will be described, and descriptions of other parts will be omitted so as not to distract from the gist of the present invention.

The main subject of the present invention is to set the initial search radius in consideration of the margin ratio and the signal received from the transmitter at the receiver, and then perform LSD on the symbols included in the set initial search radius.

To this end, the method and apparatus for LSD in the mobile communication system of the present invention will be described in detail.

1 is a mobile communication system to which the present invention is applied and shows a mobile communication system 1 having multiple access points (APs) 10a to 10c and multiple terminals 20.

Referring to FIG. 1, APs 10a to 10c communicate with terminals 20 to a base station. Each AP 10a-10c provides communication coverage for a particular cell 2a-2c. And in order to improve system capacity, the access terminal coverage area may be divided into small multiple regions, such as three sectorized cells 4a, 4b, 4c. Each of the sectorized cells 4a, 4b, 4c is serviced by a respective base transceiver subsystem (BTS). APs 10a to 10c for the sectorized cells 4a, 4b, 4c are typically co-located. The signaling transmission techniques described herein may be used for systems with sectorized cells as well as systems with non-sectorized cells.

The terminals 20 may be wireless devices, cellular telephones, personal digital assistants (PDAs), wireless modem cards, and the like, and may be connected to zero, one or multiple APs 10a to 10c via forward and reverse links at any given time. Can communicate.

And for a centralized architecture, the system controller 30 is connected to the APs 10a to 10c and controls the APs 10a to 10c. The system controller 30 may be a single network entity or a collection of network entities. For a distributed architecture, the APs 10a-10c can communicate with each other as needed.

2 shows an example of a transmitter 101 in a mobile communication system to which the present invention is applied.

Referring to FIG. 2, bits 202 are encoded via encoder 204 according to the Packet Format (PF) 208, 210 specified by rate prediction algorithm 212, and mapper 206. Symbol is mapped via The mapped symbols are then demultiplexed into M streams 216 or layers via demultiplexer 214, where M 228 is fed by receiver 226 feedback every 5 m-sec in addition to 5-bit CQI 224. It may be a designated 2-bit integer. M stream 216 is spatially mapped to modulators 220 and antennas 222 via spatial mapper 218.

3 shows an example of a receiver 226 in a mobile communication system to which the present invention is applied.

Referring to FIG. 3, the multiple antennas 232 receive a signal transmitted from the transmitter 101 via air. The output of each antenna 232 is demodulated through demodulator 233 and input to equalizer 234. The equalizer 234 may be implemented in tone units, respectively, and processes the multiple received signals into M symbols 236 for each tone. The equalizer 234 may generate information for enabling channel quality indicator (CQI) and rank calculation 231. In addition, the M symbols 236 are demultiplexed through the demultiplexer 235, and a log-likelihood ratio (LLR) for each of the demultiplexed M symbols is calculated and decoded by the detector 237. Bits 239 are output. Although not illustrated in FIG. 3, the detector 237 includes an LLR calculator for calculating the LLR and a decoder for performing the LSD.

In this case, when the detector 237 performs soft decoding, the soft decoding result may indicate an improved Bit Error Rate (BER) performance when compared to hard decision criteria. Accordingly, in order to exhibit improved BER performance in the present invention, the detector 237 performs the soft decoding. As described above, in order to decode the detector 237, an LLR of each bit is required. Thus, detector 237 calculates the LLR from a set of constant symbol vectors. In general, the reliability increases as the number of symbols required for LLR calculation increases. However, the computational complexity of the MIMO communication system is greatly increased due to high order modulation (HOM) or multiple antennas. To address this complexity, detector 237 must perform LSD using a subset of the set with all possible symbols. Accordingly, the present invention proposes a method for determining a part of a set having all possible symbols in the detector 237.

Hereinafter, after describing a method of performing LSD generally through the detector 237 of the receiver, a method of performing LSD according to an embodiment of the present invention will be described.

The MIMO communication system can be represented using Equation 1 below.

를 발생시킨다. 즉, 상기 S_c는 송신기에서의 송신 심볼 벡터이다. 여기서,

이고, 심볼 그룹은

개의 심볼로 구성된다. 따라서, 전체

개의 송신 심볼 벡터가 있다. 또한 상기 <수학식 1>에서 y_c는 수신기에서의 복소 벡터이며, n_c는 수신기에서의 잡음 벡터이다.In Equation 1, H _c is an M × M complex MIMO channel known to an observer. As described above with reference to FIG. 2, the binary signal bits are coded at a code rate R and interleaved into a binary sequence x, where x is mapped to a symbol sequence by a modulation function f.

Generates. That is, S _c is a transmission symbol vector at the transmitter. here,

The symbol group is

It consists of four symbols. Thus, full

There are two transmit symbol vectors. In Equation 1, y _c is a complex vector at the receiver, and n _c is a noise vector at the receiver.

The detector 237 generally calculates the LLR using y _c in Equation 1, i.e., a set with all possible symbols. That is, before performing LSD through the detector 237, the detector 237 changes y _c in Equation 1 to a real value y. Here the real and imaginary components are separated. The detector 237 represents the LLR in y as shown in Equation 2 below.

Here, x _{k +} and x _{k -} denote the cases where the k th bit of the sequence x is +1 and -1, respectively. If x _k is assumed to be iid (independent and identical distribution), Equation 2 may be expressed as Equation 3 below.

는

이며,

는 각 비트의 확률 질량 함수(probability mass function : pmf)이다. 그리고 상기 <수학식 3>의 우측의 제 1 항과 제 2 항은 각각 LLR의 우선적인 성분과 외부적인 성분이다. 상기 각 항은 일반적으로 L_A(x_k) 와 L_E(x_k) 로 지칭된다. 또한 상기 P(x_k) 가 초기에 동일한 것으로

가정되는 때에, L_A(x_k)=0 이다. 하지만, 외부 디코더가 이 출력을 우선적인 성분로서 MIMO 통신시스템에 피드백하는 경우, pmf는 동일하지 않게 되며

, 이후

이다.In <Equation 3>

Is

,

Is the probability mass function (pmf) of each bit. In addition, the first and second terms on the right side of Equation 3 are the preferential and external components of the LLR, respectively. Each term is generally referred to as L _A (x _k ) and L _E (x _k ). Also note that P (x _k ) is initially equal

When assumed, L _A (x _k ) = 0. However, if the external decoder feeds this output back to the MIMO communication system as a priority component, pmf will not be the same.

, after

to be.

On the other hand, in the Gaussian noise hypothesis, P (ylx _k ) in Equation 3 may be expressed as Equation 4 below.

는 k번째 비트가 x_k인 시퀀스 x로부터 맵핑된 심볼이다. In Equation 4 above

Is a symbol mapped from the sequence x where the kth bit is x _k .

Thereafter, the detector 237 may generate a soft output using the LSD as shown in Equation 5 below using the LLRs calculated from Equations 3 and 4 below.

는 k번째 LLR을 제외한 우선적인 LLR 의 시퀀스이다.In Equation 5 above

Is the sequence of preferential LLRs except the k-th LLR.

및

을 취할수록, 반복적인 디코더의 성능을 높일 수 있다. 하지만, MIMO 통신시스템에서 채널 행렬이 더 크거나 HOM이 이용되는 때에, 모든 가능한 심볼의 개수 (

)는 기하급수적으로 증가한다. 결과적으로 검출기(237)에서 상기 <수학식 5>를 이용하여 LSD를 수행하면 연산 복잡도가 높아진다.Many symbols as shown in Equation 5 above

And

The higher the number, the higher the performance of the repetitive decoder. However, when the channel matrix is larger or the HOM is used in the MIMO communication system, the number of all possible symbols (

) Increases exponentially. As a result, when the LSD is performed in the detector 237 using Equation 5, the computational complexity increases.

개의 모든 가능한 심볼 중 일부만을 이용하여 <수학식 5>를 연산하는 방안이 있다. 여기서, 리스트 ￡을 가능한 모든 심볼 중 부분집합으로 지칭한다. 상기 ￡은 ML솔루션과 낮은 메트릭

를 가지는 심볼들을 높은 확률로 포함하도록 획득한다. 여기서, 검출기(237)는 상기 심볼을 검색하기 이전에, 리스트 크기 또는 심볼 개수(N_cand)를 결정하여야 한다. 상기 ￡를 획득하기 위해 LSD를 이용하며, 상기 LSD를 이용하는 방법은 종래 SD를 이용하는 방법에 약간의 수정이 가해진 것이다.To solve this problem, the detector 237

There are ways to calculate Equation 5 using only some of all possible symbols. Here, the list ￡ is referred to as a subset of all possible symbols. The thin ML solution and low metric

Acquire symbols having a high probability to include. Here, the detector 237 must determine the list size or symbol number N _cand before searching for the symbol. LSD is used to obtain the power, and the method of using the LSD is a slight modification to the method of using the conventional SD.

First, for SD the QR division of the real value channel matrix H is H = QR, where Q and R are unitary and upper triangular matrices, respectively. Thereafter, the initial condition may be expressed as Equation 6 below.

이고,

이다. ￡의 포인트 수가 미리 정해진 심볼 개수(N_cand)로 증가할 때까지 바바이 포인트(Babai point)를 포함한 모든 검색된 심볼들이 리스트에 추가된다.Where initial radius

ego,

to be. All retrieved symbols, including Babai points, are added to the list until the number of points of _증가 increases to a predetermined number of symbols N _cand .

And while traversing the nodes in the LSD, the manner of retrieving the symbol is the same as in a typical SD. However, the search radius of a symbol in the LSD is controlled by the current list size as well as the update condition of the SD algorithm. The detector 237 does not update the symbol's search radius if it is not sufficiently filled, but adds a symbol to it that corresponds to the current point. The detector 237 compares the maximum distance metric d ₀ of ￡ with the distance metric d of the current search point when ￡ is sufficiently filled, and then returns 비교 to the current search point when the comparison result d <d ₀ . Replace the original point with d ₀ in. The detector 237 compares d with the distance metric at different points of 때에 when the replacement occurs, and updates the new search radius by the maximum distance metric.

After the search for the symbol is performed in the detector 237, Equation 5 is converted into Equation 7 below.

요소는 제로가 된다. 만일

또는

인 경우, 상기 <수학식 7>의 분자 또는 분모 각각에 매우 작은 수를 할당한다.Since Equation 7 is the first output before the external repetition, prior information is expressed by Equation 5

The element is zero. if

or

In the case of Equation 7, a very small number is assigned to each of the numerator or denominator of Equation 7.

을 갖는 많은 인수를 포함하게 하며, 이는 LLR의 신뢰성 증가를 가져온다. 그러나 복잡도의 측면에서는, 따라서, N_cand 가 클수록 복잡도 역시 크게 증가한다. 그러므로 사용자는 리스트의 적당한 크기를 결정해야 하며, 일반적으로

로 설정된다.In addition, the performance of the LSD in the detector 237 is determined by N _cand . The large N _cand is a small distance metric with high probability at ￡

It includes many arguments with, which leads to increased reliability of the LLR. But in terms of complexity, therefore, the larger N _cand , the greater the complexity. Therefore, the user must determine the appropriate size of the list, and generally

Is set to.

Accordingly, the detector 237 may reduce complexity by performing LSD using Equation 7 instead of calculating Equation 5 in the related art. However, since N _cand is fixed in Equation 7, when conditions such as signal-to-noise ratio (SNR) are large and all N _cand symbols are not needed, the detector 237 uses Equation 7. In this case, the LSD is still performed for a large number of symbols.

로의 감소를 예상할 수 있으며, 이는 직접적으로 종래보다 적게 방문될 노드 개수를 가져온다. Accordingly, the present invention proposes a method of removing a symbol having a relatively low probability of movement in both the numerator and the denominator of Equation 7 within a range that affects the entire LLR. In the method proposed by the present invention,

The reduction in furnace can be expected, which directly leads to fewer nodes to be visited than before.

To this end, an embodiment of the present invention optimally sets an initial search radius r ₀ for performing LSD as follows.

로 채워진 리스트를 획득한다. 여기서 s_i는 비용함수

에 기초하여 그 값이 작은 순서대로 정렬된다. 특히,

이다. 즉, s^*는 초기 검색 반경을 설정하기 위한 ML-솔루션(이하, 제1 포인트)이 될 수 있다. 상기 리스트 ￡는 비교적 큰

를 갖는 심볼만을 포함하는데,

는 아래의 <수학식 8>을 만족하는 것으로 가정된다.Symbol Vector Before Computing LLR at Detector 237

Get a list filled with. Where s _i is the cost function

Based on the values are sorted in small order. Especially,

to be. That is, s ^* may be an ML-solution (hereinafter referred to as a first point) for setting an initial search radius. The list ￡ is relatively large

Only include symbols with

Is assumed to satisfy Equation 8 below.

이며,

이다. 상기 <수학식 8>로부터, 양호한 성능을 위해 많은 개수의 심볼이나 비교적 작은 r_t 가 요구된다. 반면에, 보다 높은 복잡도 이득을 위해 작은 수의 심볼이나 비교적 큰 r_t가 요구된다. 이에 상기 <수학식 8>을 만족하는 심볼만을 검색하기 위해, r_t에 의해 정의되는 초기 검색 반경을 설정하는 것이 필요하다. 상기 <수학식 4>의 가정을 이용하면, 상기 <수학식 8>은 아래의 <수학식 9>와 같이 나타낼 수 있다.In Equation (8), r _t is a marginal ratio

,

to be. From Equation 8, a large number of symbols or a relatively small r _t are required for good performance. On the other hand, a smaller number of symbols or a relatively larger r _t are required for higher complexity gains. Accordingly, in order to search only symbols satisfying Equation 8, it is necessary to set an initial search radius defined by r _t . Using the assumptions of Equation 4, Equation 8 may be expressed as Equation 9 below.

In Equation (9), N is a list size corresponding to r _t and J (s *). Since all symbols in the list satisfy Equation 9, the initial search radius r ₀ is represented by Equation 10 below.

이다. 이에 따라, 본 발명의 실시 예에 따른 초기 검색 반경은

이고, 상기에서 설명한 종래 초기 검색 반경은

이므로, 본 발명의 실시 예에서 감소된 초기 반경에 따라 본 발명의 LSD는

개의 심볼을 검색할 수 있다. 또한 종래 LSD의 N_cand 가 심볼의 검색에서 고정되어 있지만, 본 발명의 실시 예에서는 LSD에서의 N은 J(s*) 또는 r_t로 인해 r₀가 각 심볼의 검색에서 고정되어 있지 않다.In Equation 10 above

to be. Accordingly, the initial search radius according to the embodiment of the present invention is

And the conventional initial search radius described above

Therefore, according to the reduced initial radius in the embodiment of the present invention LSD of the present invention is

Symbols can be retrieved. But also is fixed in the choice of a conventional LSD _cand of N symbols, N in the embodiment of the present invention LSD is not due to J (s *) or r _t r ₀ is fixed in the search of each symbol.

의 심볼들은 비용 함수에 기초하여 정렬되는 것으로 가정된다. 이후 상기 <수학식 8>에 의해,

이며, 이는 아래 <수학식 11>을 생성할 수 있다.In Equation 10, r _t may be derived as follows. As described above, the list

Are assumed to be sorted based on the cost function. Thereafter, according to Equation 8,

This can generate Equation 11 below.

로 가정하면, 이후 상기 <수학식 11>은 아래의 <수학식 12>와 같이 나타낼 수 있다.

Assume that Equation 11 can be expressed as Equation 12 below.

이고,

이다. 그리고 <수학식 12>의 우측은 g(t_i)의 표본 평균이며, 이는 대략적으로 아래 <수학식 13>과 같다.In Equation 12 above

ego,

to be. The right side of Equation 12 is the sample mean of g (t _i ), which is approximately as shown in Equation 13 below.

에서 분포된다. 현재 범위에서, <수학식 13>은 대략적으로 아래 <수학식 14>와 같다.In Equation 13, P _G (.) And P _T (.) Are pmf of G and T, respectively. And t _i is

Distributed in. In the current range, Equation 13 is roughly equivalent to Equation 14 below.

In Equation 14, f _T (t) is a probability density function (pdf) of T.

는

에서 2M의 자유도(Degree of Freedom : DOF)를 갖는 비중심적 카이-제곱 분포(chi-square distribution)

를 갖는다. 여기서,

는

의 j번째 요소이며, n은 제로-평균 가우시안 잡음이며 s_t는 s*로 근사화되는, 송신기에서 송신된 신호이다. Under the assumption of Gaussian noise,

Is

Chi-square distribution with 2M degree of freedom (DOF)

Has here,

Is

Is the j-th element of n, where n is zero-mean Gaussian noise and s _t is the signal transmitted from the transmitter, approximated to s *.

를 연산하는 것이 필요하다.

가 주로 비고정된 리스트 크기 N 에 종속하기 때문에, 정확한

를 획득하는 것은 쉽지 않다. 결과적으로, s*에 수개의 심볼을 랜덤으로 던짐으로써 랜덤 개수의 샘플(

)을 획득하고 이러한 샘플이 리스트에 있거나 s* 근처에 있는 것으로 기대하는 방식을 시도한다. 이후, s*와 이러한 샘플간의 거리 메트릭스의 평균값을 연산함으로써

를 추정하고

를 획득한다.On the other hand, the detector 237 prior to obtaining the initial search radius, non-central chi-square distribution constant

It is necessary to compute.

Is mostly dependent on the unfixed list size N,

It is not easy to get it. As a result, by randomly throwing several symbols in s *,

) And try to expect these samples to be in the list or near s *. Then, by calculating the average value of the distance matrix between s * and these samples

To estimate

Acquire it.

인 것으로 가정되지만

인 경우가 발생할 수도 있다. 이는 J(s*)가 클 경우, 특히 높은 SNR에서 발생할 수 있다. 이 경우에, 이전 송신에서 검색된 r_t의 평균값을 사용한다. 하지만, 송신이 처음 발생된 경우에,

로 정의되며, 여기서 0<D<1는 사용자에 의해 선택된다.If the upper limit of the calculated r _t is large, it is essentially

Is assumed to be

May occur. This can occur especially at high SNRs when J (s *) is large. In this case, we use the average value of r _t retrieved from the previous transmission. However, when transmission first occurs,

Where 0 <D <1 is selected by the user.

을 적분하기보다는 <수학식 14>를 연산하기 위해 효율적인 근사화를 따른다. 파트나익(Patnaik)은 비중심 카이-제곱 분포가

로 근사화될 수 있음을 제안하였다. C와 f는

와

로부터 획득되는데, 이들은

와

로서 제시된다. 여기서, 상기

는 DOF이다. 이후,

이 된다. 이러한 파라미터와 근사화에 있어서, 상기 <수학식 14>의 적분항은 아래 <수학식 15>와 같이 나타낼 수 있다.While calculating the r _t , a complex noncentral chi-square distribution term

Rather than integrating, we follow an efficient approximation to compute (14). Patnaik has a noncentral chi-square distribution

It can be approximated by. C and f are

Wow

Obtained from these,

Wow

Is presented as: Where

Is DOF. after,

Becomes In this parameter and approximation, the integral term of Equation 14 may be expressed as Equation 15 below.

의 적분보다 보다 단순하게 된다.

이고

이기 때문에, <수학식 15>의 범위

부터

까지의 적분은 제로에서 무한대까지의 적분보다 작게 되며, 이는

,

이기 때문에, 상기 <수학식 15>는 아래 <수학식 16>과 같이 나타낼 수 있다.In Equation 15 above

It is simpler than the integral of.

ego

Therefore, the range of Equation 15

from

The integral to is less than the integral from zero to infinity,

,

Therefore, Equation 15 may be expressed as Equation 16 below.

로 된다. 따라서, 마진 비율 r_t의 상한은 아래 <수학식 17>과 같다.Here, the inequality sign of Equation 12 is finally

. Therefore, the upper limit of the margin ratio r _t is expressed by Equation 17 below.

으로 설정하는 것이 바람직하다.Considering the complexity in Equation 17, r _t is the maximum value to minimize the initial radius.

It is preferable to set to.

으로 설정하여, <수학식 10>을 이용하여 초기 검색 포인트로부터의 초기 검색 반경을 설정한다. 그리고 검출기(237)는 상기 설정된 초기 검색 반경에 포함되는 심볼에 대해서 LSD를 수행한다. 즉, 본 발명에 따른 검출기(237)는 아래 도 4와 같은 과정으로 LSD를 수행한다.Therefore, after the detector 237 according to the present invention determines the first point,

By using Equation 10, the initial search radius from the initial search point is set. The detector 237 performs LSD on symbols included in the set initial search radius. That is, the detector 237 according to the present invention performs the LSD by the process as shown in FIG. 4 below.

4 shows a process for performing LSD in the detector 237 of the receiver according to the embodiment of the present invention.

로 채워진 리스트를 획득하여, 제1 포인트(s^*)를 결정한다. 그리고 검출기(237)는 상기 <수학식 8>을 만족하는 심볼만을 검색하기 위해, 마진 비율(r_t)에 의해 정의되는 초기 검색 반경(r₀)을 설정하여야 한다. 이에, 403 단계에서 검출기(237)는 <수학식 17>을 이용하여 마진 비율(r_t)를 계산한다. 그리고 405 단계에서 상기 <수학식 10>을 이용하여 상기 마진 비율(r_t)를 고려하여 초기 검색 포인트로부터 초기 검색 반경(r₀)을 설정한다. 이에 따라 407 단계에서 검출기(237)는 상기 설정된 초기 검색 반경(r₀)에 포함되는 심볼들에 대하여 LSD를 수행한다.Referring to FIG. 4, in step 401, the detector 237 performs a symbol vector before calculating the LLR.

The first point s ^* is determined by obtaining a list filled with. In addition, the detector 237 should set an initial search radius r ₀ defined by the margin ratio r _{t in} order to search only the symbols satisfying Equation (8). Thus, in step 403, the detector 237 calculates a margin ratio r _t using Equation 17. In operation 405, an initial search radius r ₀ is set from an initial search point in consideration of the margin ratio r _t using Equation 10. Accordingly, in step 407, the detector 237 performs LSD on symbols included in the set initial search radius r ₀ .

Accordingly, the present invention performs LSD by removing a symbol having a relatively low likelihood in Equation 7 (that is, by using symbols included in an initial search radius r ₀ ). It is possible to implement an LSD having a low width.

In addition, by performing the following simulation with reference to FIG. 5 according to an embodiment of the present invention, the effects on the performance and complexity of the MIMO communication system will be described.

이면, r_t를 얻기 위해

로 설정한다. 초기 반경

이고, 이에 따라 ￡=0이면, 초기반경을 계속 증가시켜 리스트가 공집합이 되지 않도록 한다.In order to perform LSD according to an embodiment of the present invention, 1000 bits and 10,000 bits are generated for an SNR of 5 dB or less and an SNR of 5 dB or more, respectively, and the 4 × 4 complex matrix H _c is a Rayleigh fading channel. Assume that Turbo coding at code rate 1/2 is then used, after which the coded sequence passes through a random interleaver. It is then assumed that 16 QAM is applied to the interleaved sequence. Also in the first transmission of all symbol vectors

, To get r _t

. Initial radius

Therefore, if ￡ = 0, the initial radius is continuously increased so that the list is not empty.

As a result of the simulation, as shown in FIG. 5, when comparing the performance difference between the LSD and the conventional LSD according to the embodiment of the present invention, the complexity difference is relatively large. That is, as the SNR in FIG. 5 increases, the number of required symbols decreases to the same N number and becomes one asymptotically. At high SNR, the average number of nodes converges to a constant value.

Meanwhile, in the detailed description of the present invention, specific embodiments have been described, but various modifications are possible without departing from the scope of the present invention. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be defined not only by the appended claims, but also by the equivalents of the claims.

Claims (10)

An apparatus for List Sphere Decoding (LSD) of a receiver in a mobile communication system,
An LLR calculator for calculating a log-likelihood ratio (LLR) for a set of symbol vectors received from a transmitter,
After calculating the margin ratio of the first point and the initial search radius based on the cost function of the symbol vector set, and after setting the initial search radius using the calculated margin ratio, the symbol included in the set initial search radius Including a decoder that performs LSD on the
The decoder calculates a cost function of each symbol vector included in the set of symbol vectors, and determines a symbol vector corresponding to the lowest cost function among the calculated cost functions as the first point. .
delete The method of claim 1, wherein the decoder,

To calculate the cost function,
Wherein J (s _i ) is the cost function, y is a real value of a complex vector at the receiver, H is a channel matrix, and s _i is an i-th symbol vector.
The method of claim 1, wherein the decoder,

To calculate the margin ratio,
Where r _t is the margin ratio, t ^* is the cost function of the first point, and

And

,

Is the noncentral chi-square distribution constant,

Is the degree of freedom of the non-central chi-square distribution.
The method of claim 1, wherein the decoder,

To calculate the initial search radius,
Wherein r ₀ is the initial search radius, J (s ^* ) is a cost function of the first point, and r _t is the margin ratio.
A method for List Sphere Decoding (LSD) of a receiver in a mobile communication system,
Calculating a margin ratio of a first point and an initial search radius based on a cost function of a set of symbol vectors input from a transmitter, and setting the initial search radius using the calculated margin ratio;
And performing LSD on symbols included in the set initial search radius.
The setting of the initial search radius may include calculating a cost function of each symbol vector included in the symbol vector set, and converting a symbol vector corresponding to the lowest cost function among the calculated cost functions into the first point. A method for LSD comprising a process of determining.
delete The method of claim 6, wherein the cost function,

Is calculated from
Wherein J (s _i ) is the cost function, y is a real value of a complex vector at the receiver, H is a channel matrix, and s _i is an i-th symbol vector.
The method of claim 6, wherein the margin ratio is

Is calculated from
Where r _t is the margin ratio, t ^* is the cost function of the first point, and

And

,

Is the noncentral chi-square distribution constant,

Is the degree of freedom of the non-central chi-square distribution.
The method of claim 6, wherein the initial search radius is,

Is set from
Wherein r ₀ is the initial search radius, J (s ^* ) is a cost function of the first point, and r _t is the margin ratio.

Images