Iterative Multiuser Decoding and Channel Estimation in DS-CDMA

  • Fleury, Bernard Henri (Projektdeltager)
  • Hu, Bin (Projektdeltager)



The division Digital Communications (DICOM) has recently developed two iterative schemes based on the EM and SAGE algorithms performing multi-user Joint data Detection and channel Estimation (JDE) in DS-CDMA systems for time-variant frequency-selective fading. Monte-Carlo simulations demonstrate the near-far resistance and the high multi-user efficiencyof the schemes, as well as their robustness towards channel estimation errors.

The objective of this project is to build up on these promising results and to identify novel, feasible iterative structures that perform Joint data Decoding and channel Estimation (JDecE) in DS-CDMA systems operating in time-variant frequency-selective channels. One of the goals of the PhD study is to modify the EM/SAGE framework to allow it for processing soft information.

The SAGE algorithm has been modified to a Soft-In Soft-Out (SISO)-SAGE algorithm that accepts soft estimates of the code symbols and provides extrinsic (EXT) information of the code symbols. In coded DS-CDMA systems, SISO-SAGE algorithm has been combined with several SISO Single User (SU) decoders to successively exchange soft information (EXT or “a posteriori” probability (APP)) of code symbols in an iterative fashion. In the resulting receiver, the SISO-SAGE algorithm performs jointly channel estimation and Multiple Access Interference (MAI) cancellation. SISO SU decoders perform symbol-by-symbol Maximum A Posterior (MAP) estimate of the code symbols.

The developed receiver architecture was implemented for turbo coded synchronous CDMA system recently. The simulations results shown in Figure 1 demonstrate some features of the JDecD receiver: i) Data decoding is robust against channel estimation errors; ii) the receiver supports a high system load (up to 1.5); iii) it is multiuser efficient. In the simulations we take random spreading sequences with spreading factor is N=8. Rate R=1/3 turbo codes with polynomials (5, 7) are used as FEC. Random interleavers are used for every 1000 information bits, equivalent to 3000 code bits. The channel is assumed to be static within 150 code bits and 6 pilot bits are employed for each 150 code bits. The system load is defined as R*K/N. K is number of users.
Effektiv start/slut dato01/10/200330/09/2006