KR101543621B1 - Apparatus and method for detecting signal in multiple input multiple output wireless communication system - Google Patents

Abstract

The present invention relates to an apparatus and method for generating candidate vectors for performing ML metrics less than a conventional MML in a multiple-input multiple-output (MIM) system, And generates a second candidate vector set (| SRLM-H, 2 |), which is a candidate vector for the vectors of the first candidate vector set, to generate a first candidate vector set And a second candidate vector generating unit for generating a final candidate vector, which is a candidate set of a small number of vectors including the ML solution, by calculating the Euclidean distance, And an Euclidean calculator for detecting the Euclidean distance.

ML detection, Euclidian distance, LLR, log likelihood ratio, candidate vector, candidate group

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to an apparatus and a method for detecting a signal of a multi-

The present invention relates to a receiving apparatus and method for a multiple-input multiple-output (MIMO) system, and more particularly, to an apparatus and method for generating a candidate vector for performing ML metrics less than a conventional MML.

2. Description of the Related Art Recently, various multimedia services in a wireless environment have been demanded due to the rapid growth of the wireless mobile communication market. In particular, a large capacity of transmission data and a high speed of data transmission are progressing. Therefore, finding a way to efficiently use limited frequencies is becoming an urgent task. In order to solve the above problem, a new transmission technique using multiple antennas is required. For example, a multiple-input multiple-output (MIMO) system using multiple antennas is used.

The multiple input multiple output system does not use additional frequency or transmission power but has an advantage of increasing channel capacity and matching with high-speed data transmission, which is a requirement of a next generation mobile communication system.

The multiple input multiple output system is a spatial diversity scheme that improves transmission reliability by obtaining a diversity gain corresponding to a product of the number of pairs of transmission and reception antennas, A spatial multiplexing scheme that increases spatial diversity, and a spatial multiplexing scheme that combines spatial diversity and spatial multiplexing.

In the case of using the spatial multiplexing scheme, a transmitter transmits different information simultaneously through each of a plurality of transmit antennas, thereby enabling high-speed data transmission. At this time, since different signals are simultaneously transmitted using a plurality of transmission antennas, a signal obtained by summing all transmission signals is received at each reception antenna of the reception terminal. Therefore, the receiver must perform a task of separating multiplexed signals for each antenna. An example of a technique for detecting a signal for each antenna at a receiving end of a system using a spatial multiplexing scheme is a technique for considering all transmittable signal vectors, and a signal vector having a minimum squared Euclidean distance . Although the ML scheme is an optimal scheme and is a standard for performance comparison with other schemes, there is a problem that it is difficult to apply to an actual system because the computational complexity exponentially increases as the number of transmission antennas and the modulation order increase .

In the receiver of the spatial multiplexing scheme, it is superior in terms of performance to transmit a soft decision value instead of conveying a hard decision value of a bit encoded by a channel decoder to decode the soft decision value It is known. Here, the input soft decision value of the decoder uses a log likelihood ratio (LLR) value as an estimated value of a modulation symbol transmitted on a channel. Therefore, the receiver of the spatial multiplexing scheme needs an algorithm for calculating an optimal LLR from a reception algorithm of low complexity as well as a reception algorithm of low complexity.

Generally, in the case of the ZF technique, which is a linear signal detection technique, the MMSE technique, and the OSIC technique, which is a non-linear signal detection technique, an operation of calculating a squared Euclidean distance is required to calculate the LLR.

However, the MML technique can find the ML solution by Euclidian calculation, but it does not consider the method of calculating the LLR of each bit.

It is an object of the present invention to provide an apparatus and method for reducing the amount of ML detection computation in a multiple input multiple output system.

It is another object of the present invention to provide an apparatus and method for reducing the area of a candidate vector set for reducing the amount of ML detection computation in a multiple input multiple output system.

According to a first aspect of the present invention, an ML detector for generating a candidate vector set for ML detection in a multiple input multiple output system includes a first candidate vector set (| A candidate symbol generator for generating a second candidate vector set (| SRLM-H, 2 |), which is a candidate vector for the vectors of the first set of candidate vectors, The Euclidean calculator detects the ML solution by calculating the Euclidean distance after generating the final candidate vector, which is a candidate set that is a prime set including the ML solution by comparing the first candidate vector and the second candidate vector. .

According to a second aspect of the present invention, an ML detector for calculating an LLR for ML detection in a multiple input multiple output system generates a first candidate vector set and a second candidate vector set for ML solution detection, A value x2 minimizing the euclidean distance of an arbitrary property store x1 and x1 minimizing the euclidean distance with respect to x2 are calculated to determine the number of times the euclidean distance is calculated An Euclidian calculator for calculating an Euclidean distance for the first candidate vector set and the second candidate vector set, and an LLR calculating unit for calculating an LLR according to the calculated Euclidean distance. do.

According to a third aspect of the present invention, there is provided a method of generating a candidate vector set for ML detection in a multiple-input multiple-output (MIMO) system, the method comprising: generating a first candidate vector set Generating a second candidate vector set (| SRLM-H, 2 |), which is a candidate vector for the generated vectors of the first candidate vector set; 1 candidate vector and the second candidate vector to generate a final candidate vector which is a candidate set that is a prime set including a ML solution and then calculate an Euclidean distance to detect the ML solution. do.

According to a fourth aspect of the present invention, there is provided a method for calculating an LLR for ML detection in a multiple input multiple output system, comprising: calculating a value x2 that minimizes the Euclidean distance of an arbitrary property store x1; Calculating x1 at which the Euclidean distance is minimized when the computed x2 is assumed to be an arbitrary property point and checking the Euclidean distance calculation count; A candidate vector set and a second candidate vector set are generated so that the elements of all the store points for the first candidate vector set | Calculating the Euclidean distance corresponding to the vector excluding the final candidate set necessary for the hard decision ML detection in all the elements of the sex store for the second candidate vector set; And calculating an LLR according to the calculated Euclidean distance.

As described above, the present invention reduces the searching space of the candidate vector set to reduce the ML detection amount in order to perform the ML metric less than the existing MML in the multiple-input multiple-output (MIM) system, Compared to existing LLR techniques that achieve optimal performance, we can also calculate the optimal LLR values from a small search space.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Reference will now be made in detail to the preferred embodiments of the present invention, examples of which are illustrated in the accompanying drawings. In the following description, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

In the following description, an apparatus and method for generating candidate vectors for performing ML metrics less than the existing MML in a Multiple-Input Multiple-Output (MIMO) system will be described.

In the following description, the ML detection technique of the hard decision detector according to the present invention is defined as RML-H (Reciprocal ML - Hard decision), and the ML detection technique of the soft decision detector is defined as Reciprocal ML - Soft decision .

1 is a block diagram showing an ML detector of a hard decision detector according to a preferred embodiment of the present invention.

Referring to FIG. 1, the ML detector may include a candidate symbol generator and an Euclidean calculator.

The candidate symbol generator of the ML detector generates a collection of candidate symbols including the ML solution.

Here, the candidate symbol generator first generates a first candidate vector set (| SRLM-H, 1 |), which is a candidate vector set for ML detection, in a manner used in a general modified ML detection technique, (| SRLM-H, 2 |), which is a candidate vector for the vectors of the first candidate vector set.

That is, the second candidate vector set becomes another candidate vector set for the first candidate vector set.

Then, the candidate symbol generator generates a final candidate vector, which is a candidate set, which is a prime number vector including the ML solution, by comparing the first candidate vector set and the second candidate vector generated in step 303. [

The Euclidean calculator detects an ML value by performing an ML metric for calculating an Euclidean distance of the final candidate vector generated by the candidate symbol generator.

2 is a block diagram showing an ML detection unit of a soft decision detector according to a preferred embodiment of the present invention.

Referring to FIG. 2, the ML detector may include a candidate symbol generator, an Euclidian calculator, and an LLR calculator.

The candidate symbol generator of the ML detector is a value that minimizes the Euclidean distance of an arbitrary property store. In other words, when x1 is an arbitrary property store, x2 and x2, which minimize the Euclidean distance, Calculate x1, which is the minimum euclidean distance, assuming a sex store.

Then, the candidate symbol generator generates a first candidate vector set and a second candidate vector set and provides the first candidate vector set and the second candidate vector set to the Euclidean calculator.

The euclidean calculator performs an Euclidian calculation on the candidate symbol received from the candidate symbol generator, that is, an ML metric process.

Here, the Euclidean calculator may perform an ML metric for the first candidate vector set by the number of times of the first candidate vector set, and for the second candidate vector set, The ML metric can be performed the number of times that the candidate vector of the hard decision detection technique is subtracted.

The LLR calculator calculates the LLR according to the ML metric performed by the Euclidean calculator.

3 is a flowchart illustrating an ML detection process of a hard decision detector according to an embodiment of the present invention.

Prior to the description of FIG. 3, the transmission vector is defined as follows for ML detection in the hard decision detector.

If an arbitrary two-dimensional complex vector (vector included in the set for ML search) satisfies Equation (1), the vector has a reciprocal if the following equation is satisfied.

Here, Q () denotes a slicing function, H denotes a channel gain matrix, and y denotes a received signal vector.

The correlation of the ML solution is defined as Equation (2) using Equation (1) defined above and a general modified ML detection equation.

Here, Q () denotes a slicing function, H denotes a channel gain matrix, and y denotes a received signal vector.

In this case, Equation (2) can be verified by using the modified ML detection formula.

Referring to FIG. 3, the hard decision detector first generates a first set of candidate vectors (| S _{RLM-H, 1} |) for ML detection in step 301. Here, the first candidate vector set is the same as the method of generating a candidate vector set for ML detection used in a general modified ML detection method. That is, the hard decision detector can generate the first candidate vector set using Equation (3) as shown in FIG.

,

이고

이다. From here,

,

ego

to be.

Then, the hard decision detector proceeds to step 303 and generates a second candidate vector set (S _{RLM-H, 2} |) that is a candidate vector for the vectors of the first candidate vector set generated in step 301.

Here, the second candidate vector set may include ML search using characteristics of only a few vectors including the ML solution among the element vectors of the MML candidate vector set according to the reciprocity of the transmission vectors Represents space.

That is, the second candidate vector set is another candidate vector set for the first candidate vector set, and the hard decision detector uses the second candidate vector set < RTI ID = 0.0 > . At this time, the second candidate vector set is 11 (| S _{RLM-H, 2} | = 11).

,

이고

이다. 여기에서, 상기 성상점 Ω'는 상기 제 1 후보 벡터 집합 생성 과정에서 얻을 수 있다.From here,

,

ego

to be. Here, the property point? 'May be obtained in the first candidate vector set generation process.

Thereafter, the hard decision detector proceeds to step 305 and compares the first candidate vector set generated in step 301 and the second candidate vector generated in step 303. [ Here, the above process is a process for reducing a search space including an ML solution to identify a candidate vector, which is a prime number vector including the ML solution.

 Then, the hard decision detector proceeds to step 307 to generate a final candidate vector according to the comparison result of step 305. Here, the final candidate vector may be generated as a candidate vector including the ML solution using Equation (5) as shown in FIG.

Here, | S _RLM-H | represents a final candidate vector, | S _{RLM-H, 1} | represents a first candidate set, | S _{RLM-H, 2} | represents a second candidate set. That is, the final candidate vector can be represented by an intersection of the first candidate vector set and the second candidate vector set, and in FIG. 4, the final candidate vector is 3 (| S _RLM-H | = 3).

In step 309, the hard decision detector performs an ML metric process to calculate the Euclidean distance of the final candidate vector set generated in step 307, and then ends the algorithm.

4 is a flowchart illustrating an ML detection process of a soft decision detector according to an exemplary embodiment of the present invention.

Referring to FIG. 4, the soft decision detector first determines a value that minimizes the Euclidean distance of any property in step 401, that is, when x1 is an arbitrary property, the Euclidean distance is minimized Is calculated using the following equation (6). Here, the establishment of Equation (6) below is proved after the description of the present invention.

Here, x denotes a transmitted symbol, y denotes a received signal vector, h denotes a channel gain, and x ₁ is included in a candidate set of the property store (x ₁ ∈Ω).

Then, the soft decision detector proceeds to step 403 and calculates x1 at which the Euclidean distance becomes minimum when Equation (2) is assumed to be an arbitrary property, using Equation (7).

Here, x represents a transmitted symbol, y represents a received signal vector, h represents a channel gain, and x ₂ is included in a candidate set of the property store (x ₂ ∈Ω).

At this time, the soft decision detector can identify the vector requiring Euclidean distance calculation from Equation (6) and Equation (7). That is, the soft decision detector

Using the assumption that Equation (6) and x1 are included in the ML search set, the following Equation (8) is calculated and Equation (7) and x2 are included in the ML search set Can be summarized as Equation (9) below.

을 구하기 위해 |Ω| 개 벡터에 해당하는 ML metric 즉, 유클리디안 거리 계산이 필요하다는 것을 의미하며 상기 <수학식 9>는

는 상기 <수학식 9>의 원소인 |Ω| 개 벡터에 해당하는 ML metric 즉, 유클리디안 거리 계산이 필요한다는 것을 의미한다.Here, Equation (8)

To obtain | Ω | Means that an ML metric corresponding to a dog vector, that is, Euclidean distance calculation is required, and Equation (9)

&Lt; / RTI >< RTI ID = 0.0 > It means that ML metric corresponding to the dog vector, that is, Euclidean distance calculation is necessary.

In step 405, the soft decision detector generates a first candidate vector set. In step 407, the soft decision detector processes the LLR for the candidate vector set generated in step 405.

Here, the soft decision detector generates a first candidate vector set for ML detection, where the first candidate vector set is the same as the method for generating a candidate vector set for ML detection used in a general modified ML detection method . That is, the hard decision detector generates the first candidate vector set using Equation (10) below.

,

이고

이다. From here,

,

ego

to be.

Also, the soft decision detector may calculate an LLR for the candidate vector using Equation (11) below.

At this time, the soft decision detector performs an ML metric corresponding to the candidate set (Ω) of the property points.

The soft decision detector proceeds to step 409 to generate a second candidate vector set, and then proceeds to step 411 to calculate the LLR for the generated candidate vector set.

Here, the soft decision detector generates the second candidate vector set using Equation (12), and calculates the LLR for the generated candidate vector using Equation (13) have.

,

이고

이다. 여기에서, 상기 성상점 Ω'는 상기 제 1 후보 벡터 집합 생성 과정에서 얻을 수 있다.From here,

,

ego

to be. Here, the property point? 'May be obtained in the first candidate vector set generation process.

In this case, since the soft decision detector has already performed the ML metric for the candidate vector set corresponding to the intersection of the first candidate vector set and the second candidate vector set, the soft decision detector The ML metric is performed as long as the candidate set is excluded.

Thereafter, the soft decision detector ends the algorithm.

The ML detection performance of the ML detector adopting the present invention is as follows.

이 도 7에 도시되었다고 가정한다.First, the environment in which the ML detector operates is shown in Table 1 below.

Is shown in Fig.

system 2x2 MIMO system (Spatial multiplexing) Channel model i.i.d. Rayleigh fading channel Channel estimation Ideal CSI at Rx Modulation method QPSK, 16-QAM, 64-QAM, 256-QAM

이 증가하고 약 15dB 이상이면 특정 값으로 수렴함을 관찰할 수 있다. 성상점 집합의 크기(|Ω|)가 증가함에 따라

값이 증가하지만, 성상점 집합의 크기를 정규화 했을 때 즉,

는 |Ω| 가의 증가에 따라 감소함을 알수 있다. 도 11에서는 다야항 성상도가 사용될때의

분포를 나타낸 도면이다.Referring to FIG. 10, as the SNR increases

And converges to a specific value when it is about 15 dB or more. As the size of the store set (| Ω |) increases

However, when the size of a property set is normalized, that is,

| Ω | And it decreases with the increase of. In Fig. 11,

FIG.

The ratio of the complexity of the existing hard decision MML scheme to the complexity of the MML scheme (RML-H) proposed in the present invention can be expressed by Equation (15) below.

As S _RML-H | increases in proportion to log ₂ | Ω | in Equation (15), the computational complexity according to the present invention decreases drastically as the constellation increases.

Further, the complexity of the LLR of the conventional soft decision detector and the method (RML-S) of the present invention can be expressed as Equation (16).

Comparing Equations (15) and (16), it can be seen that | S _RML-H | is efficient and RML-S is inefficient. However, in the case of | S _RML-H | = 1, the efficiency of the RML-S proposed by the present invention increases as | Ω | increases, as shown in Equation (16).

The comparison between the ML detector according to the present invention and the ML detector according to the conventional hard decision detector is as follows.

The ML computation volume increases in proportion to the size of the search space for ML solution detection. To compute an ML metric that computes a single euclidean distance, a norm calculation of n _R dimensional vectors is required. Since the real multiplication is required for one ML metric calculation, the computational complexity of the detection techniques can be compared as in <Table 2>. Computational complexity of the proposed RML-H in the present invention, | S _RML-H | is determined by a, of Table 2 | S _RML-H | was used the values obtained from the experimental results of the home environment.

As shown in Table 2, the complexity of the conventional ML scheme is | Ω | ² , and the existing MML increases in complexity in proportion to | Ω |. Also, it can be seen that the complexity of RML-H according to the present invention increases in proportion to approximately log ₂ |.

Table 3 compares the operation amount of the LLR generation technique of the RML-S according to the present invention with the conventional optimal LLR generation technique in terms of the number of real number multiplications.

As shown in Table 3, the amount of computation of the RML-S according to the present invention increases as the magnitude of the imaginary point collection (| Q |) increases.

The following is a value that minimizes the Euclidean distance of an arbitrary castle point in the soft decision detector according to the present invention. In other words, assuming that x1 is an arbitrary castle point, We will define the equation. Here, the equation to be proved may be Equation (6) or Equation (7) as described above.

First, to prove the above equation, a unit vector such as Equation (17) is defined, and the following Equation (18) can be obtained from the defined unit vector.

From here,

to be.

The objective function of Equation (6) can be transformed into Equation (19) as follows using Equations (17) and (18).

함수는 단조 증가함수이고, 항

은 x₂ 와 관계없는 주어진 x₁에 대해 상수임을 알 수 있다. 따라서,

를 최소화하면 되고, 이 식은 다시 하기 <수학식 20>과 같이 표현할 수 있으며 상기 <수학식 20>은 슬라이싱 함수에 따라 <수학식 21>과 같은 관계가 성립된다.In Equation 19,

The function is a monotone increasing function,

Is a constant for a given x ₁ independent of x ₂ . therefore,

, And this equation can be expressed as Equation (20) below, and Equation (20) can be expressed as Equation (21) according to the slicing function.

While the present invention has been described in connection with what is presently considered to be the most practical and preferred embodiment, it is to be understood that the invention is not limited to the disclosed embodiments, but is capable of various modifications within the scope of the invention. Therefore, the scope of the present invention should not be limited by the illustrated embodiments, but should be determined by the scope of the appended claims and equivalents thereof.

1 is a block diagram showing an ML detector of a hard decision detector according to a preferred embodiment of the present invention;

2 is a block diagram showing an ML detection unit of a soft decision detector according to a preferred embodiment of the present invention;

FIG. 3 is a flowchart showing an ML detection process of a hard decision detector according to a preferred embodiment of the present invention;

FIG. 4 is a flowchart showing an ML detection process of a soft decision detector according to a preferred embodiment of the present invention;

FIG. 5 is a diagram illustrating a first candidate vector of a hard decision detector according to a preferred embodiment of the present invention;

6 is a diagram illustrating a second candidate vector of a hard decision detector in accordance with a preferred embodiment of the present invention;

7 is a diagram showing a final candidate vector of a hard decision detector according to a preferred embodiment of the present invention;

Figure 8 illustrates a first candidate vector set of soft decision detectors in accordance with a preferred embodiment of the present invention;

9 is a graph showing simulation results of an ML detector according to a preferred embodiment of the present invention,

10 is a diagram illustrating a distribution of candidate vectors according to an ML detection technique according to a preferred embodiment of the present invention.

Claims (8)

A method for generating a candidate vector set for ML detection in a multiple input multiple output system, Generating a first candidate vector set and a second candidate vector set that are candidate vector sets for ML detection, Generating a final candidate vector set for ML detection based on the first candidate vector set and the second candidate vector set, Wherein the second candidate vector set comprises: And a candidate vector set for the first candidate vector set. delete The method according to claim 1, The final candidate vector set may include: Wherein the first candidate vector set and the second candidate vector set correspond to an intersection of the first candidate vector set and the second candidate vector set. The method according to claim 1, Wherein the generating the first candidate vector set comprises: Calculating a second property store that minimizes the euclidean distance to an arbitrary first property store; And calculating a third property store that minimizes the euclidean distance to the second property store. 1. An apparatus for generating a candidate vector set for ML detection in a multiple input multiple output system, Generates a first candidate vector set and a second candidate vector set that are candidate vector sets for ML detection and generates a final candidate vector set for ML detection based on the first candidate vector set and the second candidate vector set A candidate symbol generator, And an Euclidian calculator for performing an ML metric on the final candidate vector set, Wherein the candidate symbol generator comprises: And a second candidate vector set that is a candidate vector set for the first candidate vector set. delete 6. The method of claim 5, Wherein the candidate symbol generator comprises: And a final candidate vector set corresponding to an intersection of the first candidate vector set and the second candidate vector set. 6. The method of claim 5, Wherein the candidate symbol generator comprises: Calculating a second sex store that minimizes the euclidean distance to any first sex store, calculating a third sex store that minimizes the Euclidean distance to the second sex store, The device that generates the set.

Images