A simple soft linear detection for coded multi-input multi-output systems

Transcription

1 WIRELESS COMMUNICATIONS AND MOBILE COMPUTING Wirel. Commun. Mob. Comput. 0; :6 60 Published online October 0 in Wiley Online Library (wileyonlinelibrary.com). DOI: 0.00/wcm.0 RESEARC ARTICLE A simple soft linear detection for coded multi-input multi-output systems Youngmin Kim, Pingping Shang, Sooyoung Kim * and Kwonhue Choi Electronic and Telecommunications Research Institute, Korea Chonbuk National University, Korea Yeungnam University, Korea ABSTRACT We propose a very simple and efficient soft linear multi-input multi-output (MIMO) detection scheme. The detection process is divided into two separate problems. The proposed scheme first detects MIMO symbols using conventional linear detection methods and produces soft bit information using a simple soft demapping method. Next, we refine the soft information by accounting for uneven post-detection noise variance across MIMO layers. From the simulation result investigated in this paper, we first emphasize that powerful channel coding may suppress the differences of diversity gains among various MIMO detection schemes. This implies that the channel decoding operation may not be transparent to performance gain that resulted from MIMO detection process. The proposed scheme concentrates on accurate estimation of soft post-mimo detected information in a very simple manner, rather than concentrating on a complex MIMO detection scheme prior to decoding process. In combination with turbo codes, the proposed scheme produces comparable performance to maximum likelihood detection, even with the simplest scheme such as zero forcing detection, with drastically reduced complexity. Copyright 0 John Wiley & Sons, Ltd. KEYWORDS MIMO detection; soft demapping; turbo codes; STBC *Correspondence Sooyoung Kim, Electronics Engineering, Chonbuk National University, Korea. sookim@jbnu.ac.kr. INTRODUCTION In most future wireless systems, multi-input multi-output (MIMO) schemes will be used in conjunction with turbo codes to achieve both high spectral efficiency and power efficiency. Turbo codes are known to provide performance close to the Shannon limit, and this is achieved by an iterative soft-input soft-output (SISO) decoding process []. This excellent performance of turbo codes should be conditioned by accurate soft decision information (SDI) to the decoder input. The maximum likelihood (ML) soft MIMO detection scheme, which exhaustively searches for the SDI with the highest probability, can achieve the maximum performance. owever, the complexity increases exponentially with the number of transmitting antennas and modulation orders, and thus it is an non-deterministic polynomial time (NP) hard problem. Various heuristic approaches exist that approximately solve this problem in less than exponential time but most of them still have exhaustive estimation. The performance of turbo-coded MIMO schemes can be quite different from that of the uncoded schemes. This implies that an uncoded MIMO scheme with superior performance does not always exhibit the same behavior if it is used with a turbo code []. As an example for spatial-multiplexing (SM) schemes, lattice reduction aided detection (LRAD) schemes are considered effective in the sense that they can eliminate the noise enhancing problem of the zero forcing (ZF) detector, especially in high signal-to-noise ratio (SNR) range [][]. The superior performance of LRAD to ZF detection may not be consistent if a powerful forward error correction (FEC) coding scheme is involved. This is first because the target SNR region for the decoder is generally much lower than that for the uncoded system. The bit error rate (BER) curves of various uncoded MIMO detection schemes show a quite different trend in this lower SNR region compared with those in the nominal SNR region where we tend to design MIMO detection schemes or compare their performances. Next, the inherent characteristics of hard decision-based 6 Copyright 0 John Wiley & Sons, Ltd.

2 Y. Kim et al. A simple soft linear detection for coded MIMO systems detection for LRAD prevent it from providing accurate SDI to the turbo decoders, thus degrading the overall performance [5]. More importantly, improper application of channel information to SDI seriously degrades the decoding performance. This implies that, in a coded MIMO system requiring SDI, it is very important to estimate accurate soft input to the decoder. It was demonstrated that iterative detection and decoding with SDI could achieve near-capacity performance in the coded MIMO system [6]. owever, the complexity is too high to jointly iterate the detection and decoding process for the practical implementation. For example, the joint iterative scheme needs to invoke soft MIMO detection process at every iteration in order to update SDI. To this end, we have to memorize all the channel matrix information corresponding to a codeword size which is generally very long. This requires a large size of memory and computational time. In addition, this concatenated iterative approach requires joint optimization of detection and decoding process, which requires exponential increase in the computational complexity. In this paper, we propose an efficient linear soft MIMO detection process by dividing it into two sequential steps. The first step corresponds to conventional MIMO symbol detection, such as linear detection for space-time block coding (STBC) schemes and ZF schemes as a general case. Subsequently, the SDI for each bit in the detected symbol is extracted using a simple soft demapping process. In the next step, we refine the SDI by accounting for uneven noise variances across MIMO layers. This paper is organized as follows. Section first describes the basic background of soft MIMO detection schemes, starting with ML detection. Next, the proposed two step MIMO detection scheme with soft bit demapping is presented. Section presents application examples of the proposed soft MIMO detection for STBC and SM systems, respectively. Section demonstrates the BER performance of the turbo-coded MIMO systems using various detection schemes, compared with the uncoded MIMO systems. Finally, we draw conclusions in Section 5.. SOFT MULTI-INPUT MULTI-OUTPUT DETECTION.. Maximum likelihood detection The SDI provided to the iterative SISO decoder can be obtained using ML detection associated with log likelihood ratio (LLR), as follows []: 0PsWb L b i;j jy i;j.s/d exp jjy sjj A ; PsWb i;j.s/d0 exp jjy sjj () where the vector s is composed of n modulation symbols of s i ;i D ; :::; n, which will be estimated during a given detection interval. In addition, a modulation symbol s i is composed of m information bits of b i;j ;j D ; :::; m, where m is the modulation order. Information bits of b i; and b i;m represent most significant bit and least significant bit, respectively. ence, the vector s is composed of nm bits. is the channel matrix, is the variance of complex additive white Gaussian noise, and y is the received symbol vector. Although this ML detector may achieve the maximum performance, the complexity of computing () increases exponentially with the size of s. The following Max LLR value can be used to reduce the computational complexity but still requires exhaustive searching process to find the maximum. L b i;j jy " min jjy sjj swb i;j.s/d0 # min jjy sjj : swb i;j.s/d.. Multi-input multi-output detection detection with soft bit demapping In our proposed scheme, we first detect the transmitted modulation symbol using a conventional detection scheme. Then, we extract the soft bit information contained in each symbol using a soft demapping method. A simple soft demapping method using decision threshold values can be used [8,9]. This method does not need exhaustive Euclidean estimations but needs only single simple distance estimation. For simplicity, we first derive the SDI for M-ary quadrature amplitude modulation signals transmitted via a single antenna system. In a binary phase shift keying (BPSK) scheme where a single bit is contained in a symbol, the information bit is transmitted via in-phase (I) channel. Considering a BPSK symbol at the ith time interval, s i, only the real part of the complex modulated signal, R.s i / is transmitted. In a single antenna system with the corresponding channel gain of h i,letos i D h i y i =jjhjj denotes the channel compensated estimation of s i,whereh i denotes complex conjugate of h i,theny i D h i Os i. The LLR value for the received signal y i using ML detection in () can be derived by exp jjy i h i jj L.b i; jy i / D log exp jjy i Ch i jj D jjy i h i jj Cjjy i C h i jj () D jjh i jj R.Os i / : () From (), we can clearly see that ML detection of b i; is equivalent to the weighted distance from the hard decision threshold (DT) line, where the weighting factor is jjh i jj =, and the DT line is the imaginary axis; that is, R.y i / D 0. Extending to a quadrature phase shift keying (QPSK) scheme, in which two bits are contained Wirel. Commun. Mob. Comput. 0; : John Wiley & Sons, Ltd. 6 DOI: 0.00/wcm

3 A simple soft linear detection for coded MIMO systems Y. Kim et al. in a symbol. We can consider that each constituent bit is independent of each other and thus, the first and second bits are considered to be transmitted as independent BPSK symbols via I and quadrature-phase (Q) channels, respectively. Therefore, with the same rule, SDI for QPSK can be represented by L.b i; jy/d jjh i jj R.Os i / ;L.b i; jy/d jjh i jj I.Os i / : () Taking account of this, SDI with ML detection for each constituent bit in an M -ary symbol will be the distance from the DT line, on the condition that each bit probability is independent of each other in a symbol. This is indeed true for any conventional M -ary modulation scheme. For example, for 6-QAM with Gray mapping, the decision thresholds for four consisting bits can be represented as shown in Figure, and SDI of four consisting bits can be represented as follows: L.b i; jy i / D jjh i jj R.Os i /= ; L.b i; jy i / D jjh i jj I.Os i /= ; L.b i; jy i / D jjh i jj.jr.os i /j A/ = ; L.b i; jy i / D jjh i jj.ji.os i /j A/ = ; (5) where A denotes the lower amplitude of the I and Q channels of a 6-QAM symbol, respectively; that is, A D = p 0 for a normalized 6-QAM scheme. In Figure, Ob i;j D L.b i;jjy i / when jjh i jj D. As denoted in () (5), the SDI for each bit is the distance from the DT lines multiplied by the weighting factor of jjh i jj =.Inother words, the SDI is inverse proportional to the noise variance and proportional to the corresponding channel gain. This is eventually because the noise variance is changed due to the corresponding channel gain. owever, the aforementioned direct multiplication of channel gain may incur serious performance degradation in a MIMO system, because the noise variance across the MIMO layers will be subjected to the employed detection scheme as well as the corresponding channel gains. This implies that we need to rescale by considering the postdetection noise variance. Therefore, the final SDI input to the iterative turbo decoder in a MIMO system, b O i;j M, can be expressed by Ob M i;j D i O b i;j ; (6) where b O i;j is the SDI on an additive white Gaussian noise channel, that is, jjh i jj=. Then, i is a scaling factor accounting for the post-detection noise variance at the detected symbol Os i,thatis, i D = O i = D O i ; () where O i denotes the noise variance of Os i.inthefollowing subsections, we describe application examples of the proposed scheme to MIMO systems using STBC and SM, respectively, and derive the corresponding expressions for i s.. APPLICATIONS.. Multi-input multi-output system using space-time block coding As a simple example, we use the Alamouti scheme [0]. Assuming a flat fading channel over two time slots and a single receive antenna, the received signals are expressed by y y s D A s n C n ; (8) where y i and n i are the received signal and complex Gaussian noise added to the ith time slot, respectively. A is the channel matrix for the Alamouti code and is expressed: h h A D h h ; (9) Figure. Decision thresholds for soft demapping of 6-quadrature amplitude modulation. where h j is the channel gain of the path from the j th transmit antenna to the receiver. As A is an orthogonal matrix, multiplying its ermitian results in a diagonal matrix, and thus the detected modulation symbols, s and s can be expressed by s O D so jjh jj Cjjh jj A y y : (0) 6 Wirel. Commun. Mob. Comput. 0; : John Wiley & Sons, Ltd. DOI: 0.00/wcm

4 Y. Kim et al. A simple soft linear detection for coded MIMO systems If we insert (8) to find the post-detection noise variance, O i D then, we have so i " # " # " s O s h n C h # D C n so s jjh jj Cjjh jj h n h : n () The post-detection noise variance of Os can be represented by O h D Var n C h n jjh jj Cjjh jj : () Reminding that VarŒaX Djjajj VarŒX and VarŒXCY D VarŒX CVarŒY,whereX and Y are independent complex random variables, and a is a complex constant value, then () becomes that O D Var h n C h n jjh jj Cjjh jj D jjh jj Var Œn Cjjh jj Var n jjh jj Cjjh jj D jjh jj C jjh jj jjh jj Cjjh jj D jjh jj Cjjh jj : () It is straightforward to find that O DO. Therefore, by inserting the result of () into (), we have i for the Alamouti scheme as follows: i D = O i Djjh jj Cjjh jj : () As the second example, we consider a quasiorthogonal STBC (QO-STBC) scheme []. Assuming a flat fading channel over four time slots and a single receive antenna, the received signals are denoted as follows: y s n y 6 y 5 D s Q 6 s 5 C n 6 n 5 ; (5) y s n where Q is the quasi-orthogonal channel matrix expressed by h h h h h h h h Q D 6 h h h h 5 : (6) h h h h Considering a ZF detection scheme which is generally applied to QO-STBC schemes, the detected symbol vector Os can be expressed by Letting Os D Q Q 6 Q D s C Q Q Q Q Q D N Q D6 y y y 5 y n n 6 Q n 5 : () n Nh h N h N h N Nh h N h N h N Nh h N h N h N Nh h N h N h N 5 ; (8) where N Q becomes Q assuming a nonsingular matrix Q. Then, the noise variance after the aforementioned ZF detection for s i can be expressed as follows: X O i D Var Nh ij n j 5 j D X D Var hij N n j j D X D jjh N ij jj Var n j D j D X jjh N ij jj : (9) j D Inserting the aforementioned result into (), we have the following i for the QO-STBC scheme: i D = O i D P j D jj N h ij jj : (0) Generalizing it for any N T ZF detected QO-STBC schemes, i can be represented by i D P NT j D jj h N ij jj : () As another example of STBC schemes, we consider a QO-STBC scheme with linear detection capability (LD-QO-STBC) []. The channel matrix of the LD-QO-STBC scheme is expressed by h h h h h C h h C h L D6 h h.h h / h C h.h C h /.h h /.h h / h C h h C h 5 :.h h / h h h C h.h C h / () Wirel. Commun. Mob. Comput. 0; : John Wiley & Sons, Ltd. 65 DOI: 0.00/wcm

5 A simple soft linear detection for coded MIMO systems As L is an orthogonal matrix, L L is a diagonal matrix as follows: ˇ L LD6 0 ˇ C ˇ 0 5 ; C ˇ () where D P id jjh i jj,andˇ D R.h h C h h /. Therefore, we can detect the signal by simply multiplying the ermitian of L. M D 6 h h ::: h NT h h ::: h NT ::: ::: h NR h NR ::: h NR N T Y. Kim et al. 5 ; (8) where h ij denotes the channel gain at the path from the j th transmitting antenna to the ith receiving antenna. There are various means of performing signal detection, and we first consider ZF detection. After y is received, the ZF detected symbol vector Os can be expressed by Os D 6 ˇ ˇ Cˇ Cˇ 5 L y D s C 6 ˇ ˇ Cˇ Cˇ 5 L 6 n n ::: n 5 : () The noise variance after the aforementioned linear detection of LD-QO-STBC scheme for s i can be expressed as follows: O i D Var D " P j D.h L / # ij n j h Pj i Var D.h L / ij n j D D ; (5) where.h L / ij is the element of ith row and j th column of L, D. ˇ/ when i D ;,and D. C ˇ/ when i D ;. Inserting the aforementioned result into (), we have the following i for the LD-QO-STBC scheme: i D = O i D : (6).. Multi-input multi-output system for spatial-multiplexing We consider a MIMO system for SM with N T transmit and N R receiving antennas. Thus, the system can be denoted as follows: y D M s C n; () where s is the N T transmitting modulation symbol vector, y and n are the N R received signal and complex Gaussian noise vectors, and M is the N R N T MIMO complex channel matrix expressed by Os D M M M y D s C M M M n: (9) Letting N M D M M M,andh N ij is the element at the ith row and j th column of N M,then Os D 6 Os Os ::: Os NR C 6 5 D 6 s s ::: s NR 5 Nh h N ::: Nh h N ::: hnr N hnr N ::: ::: Nh NR Nh NR ::: hnr N N R 6 5 n n ::: n NR 5 : (0) With a similar approach to QO-STBC scheme in (9), the noise variance after ZF detection for s i can be expressed as follows: O i XN R D Var j D Nh ij n j 5 D N R X j D jj N h ij jj : () The same results was shown in [], and it was shown that O i was equivalent to the ith diagonal element of the error covariance matrix []. Therefore, i for the ZF detection of SM is expressed by i D = O i D P NR j D jj h N ij jj : () In this case, if N R D N T and M is a nonsingular matrix, then N M becomes M, and subsequently h N ij becomes the element of M. 66 Wirel. Commun. Mob. Comput. 0; : John Wiley & Sons, Ltd. DOI: 0.00/wcm

6 Y. Kim et al. A simple soft linear detection for coded MIMO systems As a second case, we consider the minimum mean square error (MMSE) detection scheme, where the symbol vector Os can be expressed by Os D M M C I M M y D M M C I M M M s C M M C I M M n: () where I M is a N T N T identity matrix. Assuming N R D N T for simplicity, and focusing on high SNR case, that is, ' 0, then the result of MMSE detection is equivalent to that of ZF detection. Therefore, we can use the same i in ().. BIT ERROR RATE PERFORMANCE SIMULATION RESULTS This section presents the performance simulation results of the turbo-coded MIMO schemes with various detection schemes. We assume that the signal is modulated using 6-QAM, and the total transmit signal power is equally divided among all the transmit antennas. We also assume that the receiver has perfect knowledge about the channel. We use a duo-binary turbo code specified as a FEC scheme in [5] with an information block size of N = symbols, that is, bits, and a code rate of /. The Max-Log-MAP algorithm was used as an iterative decoding algorithm, and the maximum iteration number was limited to eight. Figure shows the BER performance comparison of the STBC schemes over a Rayleigh fading channel. As STBC schemes, we use the Alamouti scheme in [0], the QO-STBC scheme in [], and the LD-QO- STBC scheme in []. We assume the fading is constant over two and four consecutive time periods for the and schemes,respectively. As shown in Figure, we note that the QO-STBC scheme shows the best BER performance in uncoded system, showing about db gain over the other two schemes at BER of about 0. On the other hand, the performances of the proposed soft detection method for the Alamouti and LD-QO-STBC schemes are better than that of the QO-STBC scheme, if they are combined with turbo codes. This is because the channel matrices of the Alamouti and LD-QO-STBC schemes are full orthogonal, and thus more accurate SDI can be achieved. In addition, slightly worse performance of the uncoded QO-STBC scheme in low SNR range incurs the performance degradation after the decoder, compared with those of the Alamouti and LD-QO-STBC schemes. We note that the performances of the proposed soft detection method for the Alamouti and LD-QO-STBC schemes are identical to that of ML detection. It is also very clear that the plain application of channel gain in (5), without considering post-detection noise variance, results in serious performance degradation. Although the proposed soft ZF detection for the QO-STBC shows slight performance degradation, the complexity is greatly reduced, as shown in Table I. The same complexity reduction is applicable to the turbo-coded SM schemes. Figure shows the BER performance comparison of SM scheme over a Rayleigh fading channel. We assume the fading changes at every MIMO frame. We used ZF, MMSE, LRAD, and ML detection schemes for uncoded and turbo-coded systems. In uncoded systems, the performance of LRAD is better than those of ZF and MMSE detections, regardless of the number of antennas. The diversity gain by LRAD in uncoded system becomes more evident in high SNR ranges [,]. owever, we cannot see the same gain if turbo codes are involved in MIMO systems. The output of LRAD is an inherently hard decision. For this reason, the performance of turbo-coded LRAD scheme is worse than those of the Figure. Bit error rate performance comparison of space-time block coding schemes. Wirel. Commun. Mob. Comput. 0; : John Wiley & Sons, Ltd. 6 DOI: 0.00/wcm

7 A simple soft linear detection for coded MIMO systems Y. Kim et al. Table I. Computational complexity comparison, N T D, 6-quadrature amplitude modulation. No of computations ML Proposed Matrix multiplications (s) 6 D 6 Matrix additions (y s) 6 D 6 0 exp./ 6 D 6 0 log./ D 6 0 Estimation of Ob i;j in (5) 0 D 6 ML, maximum likelihood. Figure. Bit error rate performance comparison of spatial-multiplexing schemes. proposed ZF and MMSE schemes. We also note that the performance of turbo-coded LRAD scheme is not better than that of the hard decision ZF and worse than that of the hard decision MMSE. This is quite different features from those investigated in the uncoded system. This is because of the fact that the powerful coding gain from the turbo code is mainly targeted for low SNR range, whereas diversity gain of LRAD schemes becomes dominant in high SNR range. At the SNR range where the coded system is operated, the decoder will use the uncoded MIMO detection result at the same SNR range. Because BER performances of the uncoded ZF and LRAD are almost identical Figure. Bit error rate performance comparison of turbo-coded and spatial-multiplexing schemes. 68 Wirel. Commun. Mob. Comput. 0; : John Wiley & Sons, Ltd. DOI: 0.00/wcm

8 Y. Kim et al. A simple soft linear detection for coded MIMO systems at that low SNR range, the decoder also produces almost identical decoding performances for both bard decision ZF and LRAD. Compared with ML detection, the proposed soft ZF detection exhibits db power loss with dramatically reduced detection complexity. Comparing with hard decision detection, the proposed scheme shows about a 5 db gain. Similar to STBC cases, it is also very clear that the plain application of ZF detection as in (5), without considering post detection noise variance, results in serious performance degradation. In this case, the performance degradation is even more serious in the case of STBC schemes. Suitable interleavers operating between codewords can increase the performance, as reported in [6], because the MIMO scheme produces the bulk of channel information for a MIMO frame consisting of multiple bits; thus reducing the time diversity effect of the FEC schemes. Figure compares turbo coded and SM with various detection schemes as well as a bit interleaving (BI) scheme. Compared with a coded system with BPSK producing N different fading information across the codeword of length N,anN T N R system with M -QAM will produce N=L fading information, where L=N T log M is the size of a MIMO frame in bits. Thus, we set the BI depth to 8 and 6 for and system with 6-QAM, respectively. Compared with ML detection, the proposed soft ZF detection exhibits and db power loss with dramatically reduced detection complexity for and MIMO systems, respectively. Although the performance of soft ZF detection is worse than the soft ML performance, we can achieve superior performance in combination with BI. It is obvious that even with BI, soft ZF detection has much lower complexity than ML detection without BI at the expense of interleaving delay. 5. CONCLUSION In this paper, we presented an efficient soft linear detection method for turbo-coded MIMO schemes. From the simulation results shown in this paper, the improper application of soft MIMO detection value to iterative decoder, without considering post-mimo detection noise variance, results in serious performance degradation. The simulation results demonstrated that turbo coded MIMO detection performance shows quite different features from those investigated in the uncoded system. This is because the powerful coding gain from turbo codes is mainly targeted for low SNR region, whereas diversity gain of MIMO detection schemes appear in high SNR region. This implies that in turbo-coded systems, complex MIMO detection may be useless unless accurate SDI estimation is not provided. The proposed soft MIMO detection for STBC could provide the ML performance. Although, compared with ML detection, the proposed soft ZF detection exhibits and db power loss for and SM schemes systems, respectively; the computational complexity is dramatically reduced. The proposed soft linear MIMO detection with bit interleavers produces comparable performance to ML detection without interleaver with much less computational complexity. ACKNOWLEDGEMENT This research was supported by the Basic Science Research Program through the NRF, Korea, funded by the Ministry of Education, Science and Technology (grant number ). REFERENCES. Berrou C, Glavieux A, Thitimajshima P. Near Shannon limit error-correcting coding and decoding: Turbo-codes, In IEEE Int. Conf. Commun, Geneva, Switzerland, 99; Park U, Kim Y, Kim S. A new result on turbo coded QO-STBC schemes. IEEE Communications Letters 00; (): Yao, Wornell GW. Lattice-reduction-aided detectors for MIMO communication systems. IEEE Globecom 00 00; : 8.. Berenguer I, Wang X. MIMO antenna selection with lattice-reduction-aided linear receivers. IEEE Transactions on Vehicular Technology 00; 5(5). 5. Shang P, Kim Y, Kim S, Choi K. Consideration of soft MIMO detection for turbo codes. In Proceedings of the 6th Int. Conf. on Wireless Communications, Networking and Mobile Computing, ochwald B, Brink ST. Achieving near-capacity on a multiple antenna channel. IEEE Transactions on Communications 00; 5: Larsson EG, Jalden J. Fixed-complexity soft MIMO detection via partial marginalization. IEEE Transactions on Signal Processing 008; 56(8): Tosato F, Bisaglia P. Simplified soft-output demapper for binary interleaved COFDM with application to IPERLAN/. IEEE International Conference on Communications 00; : Ryoo S, Kim S, Lee SP. Efficient soft demapping method for high order modulation schemes, In CDMA International Conference 00, Seoul, Korea. 0. Alamouti SM. A simple transmit diversity technique for wireless communications. IEEE Journal on Selected Areas in Communications 998; 6(8): Jafarkhani. A quasi-orthogonal space-time block code. IEEE Transactions on Communications 00; 9():. Wirel. Commun. Mob. Comput. 0; : John Wiley & Sons, Ltd. 69 DOI: 0.00/wcm

9 A simple soft linear detection for coded MIMO systems Y. Kim et al.. Park U, Kim S, Lim K, Li J. A novel QO-STBC scheme with linear decoding for three and four transmit antennas. IEEE Communications Letters 008; (): Ru C, Yin L, Lu J. An LDPC coded MIMO-OFDM system with simple detection an channel estimation scheme, In IEEE 6th CAS Symp. on Emerging Technologies, China, May - June, 00.. Wübben D, Kammeyer K. Low complexity successive interference cancellation for MIMO-OFDM systems. European Transactions on Telecommunications 00; 8: IEEE Std IEEE standard for local and metropolitan area networks, Part 6: Air interface for broadband wireless access systems, Giresset N, Brunel L, Boutros JJ. Space-time coding techniques with bit-interleaved coded modulation for MIMO block-fading channels. IEEE Transactions on Information Theory 008; 5(5): AUTORS BIOGRAPIES Youngmin Kim received the BS and MS degrees from Chonbuk National University in 009 and 0, respectively. e is now a member of the engineering staff in terrestrial broadcasting technology research team, Electronics and Telecommunications Research Institute (ETRI), Korea. is research interests include MIMO schemes for terrestrial broadcasting system. Pingping Shang received the BE degree from South-Central University for Nationalities, China, and MS degree from Chonbuk National University, Korea in 008 and 00, respectively. She is currently pursuing her PhD degree at Chonbuk National University. er research interests include channel coding theory and mobile communications. Sooyoung Kim received the BS degree in electrical and electronics engineering from KAIST, Korea in 990. After having worked at Satellite Communication Technology Division, ETRI, Korea from February 990 to September 99, she received the MSc and PhD degree in Electrical and Electronics Engineering from the University of Surrey, UK in 99 and 995, respectively. From November 99 to June 996, she was employed as a research fellow at the Centre for Satellite Engineering Research, University of Surrey, UK. In 996, she rejoined the Satellite Communication Technology Division, ETRI, Korea and worked as a team leader until February 00. She is now an associate professor in Chonbuk National University, Korea. er research interests include forward error correction coding schemes and multicarrier transmission techniques. She has been working on ITU-R since 000, and now, she is actively working on Working Party B of ITU-R. She has been a technical chair and guest editor at various conferences and international journals, including satellite systems track at VTC 008 spring and International Journal of Satellite Communications and Networking. Kwonhue Choi received the BS, MS, and PhD degrees in Electronic and Electrical Engineering from Pohang University of Science and Technology (POSTEC), Korea in 99, 996, and 000, respectively. From 000 to 00, he was with Electronics and Telecommunications Research Institute (ETRI), Korea as a senior research staff and worked for the development of efficient transmission algorithms for satellite communications. e was a visiting professor at Oregon State University, USA from September 008 to October 009. e is currently with Yeungnam University, Gyongsan, Korea as an associate professor. is research interests include performance analysis of wireless communication systems and digital modem algorithm design. Currently, he is interested in the efficient multiple access, diversity schemes for wireless fading channel environment, cooperative communications, and complexity reduced MIMO transceiver design. 60 Wirel. Commun. Mob. Comput. 0; : John Wiley & Sons, Ltd. DOI: 0.00/wcm