Multiuser detection with joint channel estimation for asynchronous CDMA flat Rayleigh channels

In our previous work with asynchronous CDMA flat Rayleigh fading channels, multiuser detection was coupled with a disjoint channel estimation approach where each user employed a separate Kalman filter to estimate its complex-valued fading coefficients. In this paper, we turn our attention towards joint estimation where a single Kalman filter is utilized to estimate the channel coefficients of all users in the system. Due to the high dimension of the Kalman filter, there is an increased computational cost associated with joint estimation. However, the primary interest here is in evaluating any performance gains that may be achieved from a joint estimation strategy. In this paper, we consider two joint estimation strategies: decision-directed and stochastic. For multiuser detection, the Kalman channel estimators are combined with either the decorrelating decision feedback (DDF) detector or the probabilistic data association (PDA) detector. Simulation results show that in most cases, the performance of the joint schemes are comparable to the disjoint approach; however, when the system is overloaded, the joint schemes provide improved performance.