2006 IEEE Ninth International Symposium on Spread Spectrum Techniques and Applications Multiluser Decorrelat'ing Based Long-Range Frequency-Domailn Channel Transfer Funct'ion Prediction in Multicarrier DS-CDMA Systems Bin Hu, Wei Liu, Lie-Liang Yang and Lajos Hanzo School of ECS, University of Southampton, S017 1BJ, UK. Tel: +44-23-8059 3125, Fax: +44-23-8059 4508 Email: bh2O2r,wlO3r,ily,lh @ ecs.soton.ac.uk; http://www-mobile.ecs.soton.ac.uk Abstract- In this contribution, multiuser decorrelating based CTF prediction. In [7], the CIR was predicted based on the Minimum frequency-domain channel estimation and long range channel prediction Mean Square Error (MMSE) estimation principles. The most impor- techniques are proposed for a generalized Multicarrier DS-CDMA (MC tant characteristic of this algorithm is that the sampling rate is typically DS-CDMA) system communicating over a fast fading and frequency- . . a selective channel. In the MC DS-CDMA system considered, the channelROOT-MUSIC algo- transfer function (CTF) is estimated in the frequency-domain with the aid rithm was invoked for non-dispersive channel envelope prediction. By of pilot symbols by invoking the multiuser decorrelating based channel contrast, the ESPRIT algorithm was employed for the prediction of estimation technique, in order to reduce the effects of both the Multiuser fast-fading wideband channels in [9]. Furthermore, in [I0], r 1 ] both Interference (MUT) and the background noise. Then, Kalman filter as- . o sisted long-range channel prediction is carried out with the aid of both the te em ension e-Diand 2-D U ni eSPRIT a s hae current and previous frequency-domain CTFs for the sake of predicting been employed for estimating the CTF. Once the CTF has been de- the future CTFs. Furthermore, a sinc-interpolator is employed for the termined, its future values can be extrapolated in both the time and sake of deriving the frequency-domain CTFs associated with the data frequency domain using the techniques proposed in Chapters 15 and symbols. Our simulation results show that for a reasonable signal-to-noise 16 of [2] ratio (SNR) value the proposed frequency-domain multiuser decorrelating based channel estimator is robust to the effects of both the MUI as well asnvoke the multiuser the noise. decorrelating based channel estimation scheme in the context of gener- alized MC DS-CDMA jfr the sake of estimating the frequency-domain (FD) CTF while reducing the effects of both the Multiuser Interfrrence I. INTRODUCTION (MUI) and the background noise. Then, Kalman filter assisted long One of the most challenging problems in high data rate wireless sys- range FDCTF prediction is carried out in order to predict thefuture tems is that of overcoming the effects of dispersion imposed by mul- CTFs based on both the current and previous CTFs determined with tipath propagation. Based on a combination of Direct Sequence Code the aid of the dedicated MC DS-CDMA pilot symbols to be described Division Multiple Access (DS-CDMA) and Orthogonal Frequency Di- in Section III. Finally, we generate thefirequency-domain CTFs asso- vision Multiplexing (OFDM)- MC DS-CDMA [1] [2] [3] [4] [5 ciated with the data symbols by employing a sinc interpolator has been proposed for a variety of high-rate wireless communication The rest of this paper is organized as follows. In Section II the applications. In this contribution , we discuss the generalized MC DS- philosophy of the uplink generalized Multicarrier DS-CDMA system CDMA system investigated in [6], which incorporates the subclasses and the wideband wireless channel are briefly described. In Section of both multitone DS-CDMA [3] and orthogonal MC DS-CDMA [4] III the multiuser decorrelating based channel estimation scheme is in- as special cases. In the MC DS-CDMA system considered, the en- vestigated while the long range channel predictor assisted by both a tire frequency band is divided into a number of subcarriers. Signals Kalman filter and a sinc-interpolator are considered in Section IV. The transmitted in each subcarrier experience flat fading, provided that the attainable performance is studied in Section V. Finally, Section VI of- bandwidth of each subcarrier is lower than the coherence bandwidth of fers our conclusions. the channel. Moreover, a sufficient long cyclic prefix can be incorpo- rated for the sake of compensating for both the asynchronous delay dif- 11. SYSTEM MODELS ferences of the different users as well as for the delay-spread-induced inter-symbol interference (ISI) imposed by the dispersive channel [2]. In this case, each of the subcarriers can be estimated or predicted using - IFFT l a variety of schemes designed for flat fading channels. b,(t) k '2 On the other hand, in future wireless systems, the carrier frequency Ck(t) cos(27ft) is likely to be high, which results in normalized high Doppler frequen- k cies. Thus, using the outdated channel transfer function (CTF) esti- Fig. 1. Transmitter schematic of MC DS-CDMA using both time-domain and mated based on the past received data using Decision Directed Chan- frequency-domain spreading nel Estimation (DDCE) principles [2] may not be sufficiently accurate. However with the aid of long range CTF prediction the future Chan- The transmitter of the generalized MC DS-CDMA system is por- nel T"Impulse Repo"nseo (CIR)Tor CTFmaybeestimatedsufficientlyac- trayed in Fig. 1L. At the transmitter side the binary data stream bT (t) iS curately [7]. Various algorithms have been proposed inl the literature speduiga -ciiedmi DS spedn waefr c. (t. [7],[8][9][10 [1] fr te ske o imlemntig lng ang CI or The DS spread siLgnals are simultaneously modulated usiLng Bilnary Thefinncil sppot o th Euopan nio uner he uspcesof he Phase Shift Keyinlg (BPSK) and then spread usinlg a frequency do- PhoenLix alnd New omn proje ts anld that of the EPSlRC,UK is gratefully ac- main1 orthogolnal spreading sequelnce c'kc [Ck s1, Ck 2,2... ,: Ck ZzV of knowledged. length V, where we have c'kc c' k 1. Inl our inlvestigationls we 0-7803-9780-0/06/$20.00 ©)2006 IEEE 163