9 Multi-User MIMO Receiver Processing for Time-Varying Channels

Wireless broadband communications for mobile users at vehicular speed is the cornerstone of future 4th generation systems. This chapter deals with joint iterative channel estimation and multi-user detection for the uplink of a multi-carrier (MC) code division multiple access (CDMA) system. MC-CDMA is based on orthogonal frequency division multiplexing (OFDM) and employs spreading sequences in the frequency domain [Kai98]. Both the mobile stations and the base station employ multiple antennas, hence we deal with a multi-user multiple-input multiple-output (MIMO) receiver. So far, most research on multi-user detection has dealt with block-fading frequencyselective channels, where the channel state is assumed to stay constant for the duration of a single data block of K data symbols. Even so, the optimal maximum a-posteriori (MAP) detector for such a system is prohibitively complex although it can be approximated using iterative linear minimum mean-square error (LMMSE) multi-user detection and parallel interference cancelation [ZMWM06]. This work deals with mobile users where the MIMO channels are timeand frequencyselective. Due to the rapid time-variation of the MIMO channel, the computational complexity of conventional multi-user receivers, based on channel estimation, parallel interference cancelation, multi-user detection and iterative decoding, increases drastically since the multi-user detection filters need to be re-calculated for each data symbol individually. In this chapter, we address this complexity issue by trading accuracy for efficiency. As a starting point, we adopt a joint iterative structure based on LMMSE multi-user detection and channel estimation. The decoding stage, implemented by the BCJR algorithm [BCJR74], supplies extrinsic probabilities (EXT) and a-posteriori probabilities (APP) on the code symbols. This APP and EXT information is fed back for enhanced channel estimation and multi-user detection, respectively [ZMWM06, MWZ+06]. The remainder of this chapter is organized as follows: In Section 9.2, the signal model is established and in Section 9.3, key ideas for complexity reduction are introduced. For complexity reduction, we