Frequency domain estimation of time varying channels in OFDMA systems: An EM approach

OFDM modulation combines advantages of high achievable data rates and relatively easy implementation. However, for proper recovery of input, the OFDM receiver needs accurate channel information. Most algorithms proposed in literature perform channel estimation in time domain which increases computational complexity in multi-access situations where the user is only interested in part of the spectrum. In this paper, we propose a frequency domain algorithm for channel estimation in OFDMA systems. The algorithm performs eigenvalue decomposition of channel autocorrelation matrix and approximates channel frequency response seen by each user using the first few dominant eigenvectors. In a time variant environment, we derive a state space model for the evolution of the eigenmodes that help us to track them. This is done using a forward backward Kalman filter. The performance of the algorithm is further improved by employing a data-aided approach (based on expectation maximization).