Sunteți pe pagina 1din 4

A Low Complex V-BLAST Detection Algorithm for MIMO-OFDM System

Wang Bei
Jiangsu Key Laboratory of Wireless Communications Nanjing University of Posts and Telecommunications Nanjing, P.R.China kukuabb@hotmail.com

Zhu Qi
Jiangsu Key Laboratory of Wireless Communications Nanjing University of Posts and Telecommunications Nanjing, P.R.China zhuqi@njupt.edu.cn which have different bit error performances and complexities. ML-DFE algorithm increases the former l sub-layers performance with a relatively lower complexity, thereby improves the overall system bit error performance. However, QR decomposition, finding unitary matrix, matrix decomposition etc., all of these are the extra additional matrix calculation. OSIC-ML algorithm presented in this paper, as the simplified algorithm of the ML-DFE, by exchanging the executive order of OSIC and ML, and using simple matrix operations to replace the method of through QR decomposition to find a unitary matrix for matrix resolution, decreases complexity of the matrix operations, and gets almost the same performance of ML-DFE. II. SYSTEM MODEL

AbstractML-DFE algorithm is known as the combination of the ZF-DFE algorithm and ML algorithm. In this paper, OSIC-ML is proposed as the simplified algorithm of the ML-DFE, by exchanging the executive order of OSIC and ML, using simple matrix operations to replace the method of through QR decomposition to find a unitary matrix for matrix resolution. ML-DFE algorithm decreases complexity of the matrix operations, and gets almost the same performance of ML-DFE algorithm by the Matlab simulation results. Keywords- MIMO-OFDM; Algorithm; ML; OSIC V-BLAST; Signal Detection;

I. INTRODUCTION MIMO (Multiple-Input Multiple-Out-put) is known as a big breakthrough in smart antenna techniques of wireless mobile communication areas. It can enhance the spectral efficiency several times without increasing bandwidth. The combination of MIMO and OFDM (Orthogonal Frequency Division Multiplexing) are the key techniques of 4G for their higher spectral efficiency and the ability to combat multi-path fading. MIMO is comprised of transmit diversity and spatial multiplexing. The transmit diversity contains the same information of symbols on different antennas and achieves spatial diversity. Spatial multiplexing is different from transmit diversity, it contains the different information on the different antennas, and it improves the capacity of MIMO in nature. Layered space-time is a typical application of spatial multiplexing, especially V-BLAST (Vertical Bell Laboratories Layered Space-Time coding), has broad application prospects in the MIMO-OFDM system of signal detection technology. Applied V-BLAST decoding algorithm to signal detection, ML (Maximum Likelihood) algorithm [1] is the best way. However, as the complexity of algorithm exponentially growing by the number of transmission antennas, it is difficult to apply the knowledge to actual. The OISC (Ordered Successive Interference Cancellation) algorithm, which is proposed in reference [2], needs to eliminate the impact of the element after its detection, after then allows the next element detected in a pre-determined order. In short, this algorithm sacrifices the performance in exchange for the reduction of the complexity. ML-DFE algorithm [3] is an effective classical algorithm of combining ZF-DFE algorithm and ML algorithm
This work is supported by 973 Program (2007CB310607) National Natural Science Foundation of China (60772062) and Jiangsu Provincial Science Technology Project BG2006039

A. Notations Vectors are denoted by underlined letters. Other notations used are as follows.

(.) H (.)T (.)

Hermitian of

(.) . Transpose of (.) . 2-norm of (.) .

B. System Model The transmitting end and the receiving end of the M TxN Rx MIMO-OFDM system block diagram using V-BLAST algorithm for signal detection are illustrated in Figure 1and Figure 2. The transmission sequences {X m }m =1, 2....M through IFFT transform, cyclic prefix inserting (+ CP), are sent by the

m th

form X m

Y deleting CP, FFT transform, the sequences { n }n =1, 2.... N are


detected by V-BLAST algorithm, thus the original sequences

~ {Y }

{~ }

( m = 1,2....M ) transmitting antenna as the sequences


m =1, 2....M

. At the receiving end, the signals


th

n n =1, 2.... N

are received by the n

( n = 1, 2... N ) antenna, after

{X }

m m =1, 2... M

are restored.

978-1-4244-2108-4/08/$25.00 2008 IEEE

Extract one symbol from every X m and Yn for instance. The N -dimensional received signal vectors Y can be expressed mathematically as:

~ ~ Y =HX +N

(1)

Where X demotes the M -dimensional transmit symbol vectors, H = ( hij ) N M is the N M channel gain matrix, whose element hij at the i
th th

row and j column is the


th

th

channel gain form the i transmit antenna to the j receive antenna, and they are assumed to be independent and identically distributed (i.i.d.) complex Gaussian random variables with zero mean and unit variance. The elements of the N -dimensional noise vector N are assumed to be i.i.d. complex Gaussian random variables with zero mean and variance of 2 . It is verified that the V-BLAST detection for the entire MIMO-OFDM system could be achieved by using V-BALST detection in every sub-carrier channel [4]. The number of subcarriers are usually admitted for the K = 256 for MIMOOFDM system, the V-BLAST detection of MIMO-OFDM system is equivalent to 256 M N channels for V-BLAST detection combined.

decomposition of the channel matrix H ,so as to eliminate the coupling of the remaining stratification of the former l ( l degrees of Maximum Likelihood ,the number of sub-layers for ML decoding) sub-layers to the rest, then applies ML algorithm to the l sub-layers, after a series process of detection and elimination, does the decision feedback processing on the following sub-layers. ML-DFE algorithm increases the former l sub-layers performance with a relatively lower complexity, thereby improves the overall system bit error performance. However, the decision-feedback structure still exists between sub-layers; the proliferation of error is unavoidable. The full ML-DFE[3] detection algorithm can be described as follows:

H = (Qa Qb )

Ha Hb

(2)

Where Qa and R are get by the QR decomposition of the former ( M l ) rows of the channel matrix H . Qa is an unitary matrix of size N ( M l ) , R is an upper triangular matrix of size ( M l ) ( M l ) . Now to find an arbitrary Qb of size N l such that

X1 X2

~ X1 ~ X2
~ XM

Qb Qa = 0 and Qb Qb = 1 .Then choose H a and H b


H H

such that

Ha Hb

= (Qa , Qb ) (hM l +1 , hM l + 2 , ..., hM )


H

XM

Figure 1.

The transmiting end of V-BLAST MIMO-OFDM system

(3) Where H a a matrix of size ( M l ) l , H b is a square matrix of size l l . Using this decomposition, first to detect T ( x N l +1 , x N l + 2 , , x N ) jointly by ML decoding of size l , cancel the interferences caused by these symbols, and then T detect ( x1 , x 2 , , x N l ) by the usual DFE procedure.

~ Y1 ~ Y2

Y1 Y2

X1 X2
XM

~ YN
Figure 2.

YN

The receiving end of V-BLAST MIMO-OFDM system

B. The Proposed Detection Algorithm (OSIC-ML) Processing ML detection prior to OSIC detection, interference suppression is required to deal with first, so that an additional matrix operation is added. If we exchange the executing order of the two detection algorithms, try to maintain the performance as much as possible, as well as to reduce the complexity under certain conditions, this kind of improved detection algorithm proposed in this paper--OSICML, will be more effective and practical. The details of OSIC-ML algorithm are as follows:
[3]

III.

THE PROPOSED DETECTION SCHEME

A. The Origial Detection Algorithm of V-BLAST(ML-DFE) ML-DFE algorithm is known as the combination of the ZF-DFE algorithm and ML algorithm. It does partly QR

1. Take MMSE-OSIC algorithm [5] as a representative for OSIC. Detect aki by MMSE-OSIC, marked the corresponding ( H ) ki :

978-1-4244-2108-4/08/$25.00 2008 IEEE

i 1 r1 = r G1 = ( H H H + 2 I M M ) 1 H H

k1 = arg min (G1 ) j ,


j

While ( ( M l ) )

75 km / h, and the carrier frequency is 2 GHz. At the receiving end of the signal detection, the impact of channel estimation is not considered for assuming that the information of channel is well known. Figure 3 shows the performance comparison of ML-DFE, OSIC-ML in the condition of SNR form 0dB to 20dB, with l = 2 and l = 3, as well as ML algorithm.

wki = (Gi ) ki , y ki = wki ri , a ki = Q ( y ki ) ,


T

ri +1 = ri a ki ( H ) ki ,

Gi +1 = ( H ki H ki + 2 I M M ) 1 H ki ,
H H

k i = arg min
2. A = [ a k1 , a k 2 ,

j{k1 ,..., k i 1 }

(Gi ) j , i i + 1

a k M l ]T , split the channel matrix H into , ( H ) kM l ] , (H ) kM ]


Figure 3. Performance comparison for OSIC-ML,ML-DFE and ML

H a and H b :
H a = [( H ) k1 , ( H ) k2 , H b = [( H ) kM L +1 , ( H ) k M l + 2 ,

Where k M l +1 ,....., k M {k1 , k 2 ,...k M l } 3. Eliminate the impact of the detected vectors:

~ = r H A r a
4. Detect the rest l sub-layers signals by ML detection:

a k M l +1 ......k M = arg min

r 'k M l +1 ...... k M E

~ H r ' r b k M l +1 ...... k M ,

Where k M l +1 ,....., k M {k1 , k 2 ,...k M l } 5. According to the value of k i , resume the receiving vector

[a1 , a 2 ,....a M ] .
As the basis of ML-DFE , at goal of complexity reduction, OSIC-ML algorithm by exchanging the executive order of OSIC and ML, using simple matrix operations to replace the method of through QR decomposition to find a unitary matrix for matrix resolution, decreases the amount of computation of H a and H b comparing to ML-DFE algorithm. However, due to the executive order of OSIC and ML of OSIC-ML algorithm, there will be some kind of error diffusion existed. But according to the simulation results, the slight reduction in performance could be accepted. Therefore, as the ML-DFE simplified algorithm, OSIC-ML algorithm is practicable. IV. SIMULATION AND RESULTS

ML-DFE algorithm and ML-OSIC algorithm have high flexibility because of the variety of the parameter l . The performance can be balanced with the complexity by choosing different l , even could be equivalent to OSIC algorithm or ML algorithm. The performances of ML-DFE and OSIC-ML are more close to ML algorithm along with the increase of l . When l = 3, the performances of algorithms are better than that in the condition of l = 2 . In the same l , ML-DFE algorithm is slightly better than OSIC-ML, which is because the continuous detection way used in OSIC, the bit error performance of the first sub-layers will affect the following judgments, so that the OSIC- ML algorithm has a relatively poorer performance than ML-DFE algorithm.

The simulation system preferred in this paper is a 4Tx-6Rx V-BLAST MIMO-OFDM system. The length of OFDM symbol is 256, CP is 64, QPSK modulation. The channel model is M.1225 recommended by ITU-R. The sampling frequency is f s = 4 MHz , the mobile station moving speed is

Figure 4. Performance comparison for 4Tx-6Rx and 4Tx-8Rx

978-1-4244-2108-4/08/$25.00 2008 IEEE

Figure 4 shows the performance comparison for 4Tx6Rx and 4Tx-8Rx with l = 2 . The performance will be greatly improved with the increasing of the receiving antenna, while transmitting antennas are the same. Because the diversity gain is greater with the receiving antenna increasing. It is generally considered that the more receiving antenna is applied, the better performance is get, but also the complexity of calculation is correspondingly rising.

V. CONCLUSION OSIC-ML simplifies the complexity of algorithm, as well as owns almost the same bit error performance to the ML-DFE. According to the requirements of different systems, different degrees of Maximum Likelihood ( l ) could be choose to meet the system performances and complexity requirements. But they are still unable to access the best performance of ML algorithm; therefore, the technology of V-BLAST detection algorithm could be more explored. There are some entrances to combined STBC to add diversity gain, or to develop a novel algorithm to replace the ML with the lower complexity and the better performance, which will be the directions of the further research. REFERENCES
[1] R.Van Nee, A.van Zelst, G. Awater, Maximum likelihood decoding in a space division multiplexing system, Proc. of IEEE Vehicular Technology Conference (VTC) 2000. Tokyo, vol. 1, no.5, pp. 6-10, 2000 P.W. Wolniansky, G.J.Foschini G J and G.D. Golden etal, V-BLAST: An Architecture for Realizing Very High Data Rates Over the RichScattering Wireless Channel, URSI International Symposium on Signals, Systems, and Electronics (ISSSE98) C . Italy: Pisa, pp. 295300, Sept.-Oct., 1998. Won-Joon Choi, Rohit Negi, and John M.Cioffi, Combined ML and DFE decoding for the V-BLAST System, Proc.IEEE ICC'00[C], pp. 1243-1248, 2000. L Giangaspero, L Agarossi, G Paltenghi, Co-channel interference cancellation based on MIMO-OFDM systems, IEEE Wireless Communications, vol..9,no.6, pp.8-17, 2002. D. Wbben, R. Bhnke and J. Rinas etal, Efficient algorithm for decoding Layered Space-Time Codes, IEE Electronic Letters, vol.37, no.22, pp. 1348-1350, 2001. ZHENG Shan-shan, CHEN Xia, TAN Zhen-hu, Detection Algorithms for V-BLAST Systems, Wireless Communication Technology, vol..4, pp.23-27, 2006.

[2]

[3]

Figure 5. Performance comparison for QPSK and 16QAM

[4]

Figure 5 shows the performance comparison for QPSK modulation and 16QAM modulation. The bit error performance of the former gets about 3 dB increases than the latter.

[5]

[6]

978-1-4244-2108-4/08/$25.00 2008 IEEE

S-ar putea să vă placă și