EM-MMSE BASED CHANNEL ESTIMATION FOR OFDM SYSTEMS

Abstract

Orthogonal frequency division multiplexing (OFDM) is an effective way to increase data rate and simplify the equalization problem in wireless communications. OFDM systems require an efficient channel estimation procedure to demodulate the received data coherently. Accurate estimation of channel parameters is a difficult task in OFDM systems. Pilot based techniques have normally been employed for channel estimation in OFDM systems. The obvious drawback associated with pilot based techniques for channel estimation is bandwidth overhead. To reduce the bandwidth overhead, EM algorithm is an improved way to estimate the channel coefficients in OFDM systems. The EM algorithm is a semi-blind iterative estimation procedure with expectation and maximization steps and it requires initial estimation. An initial -estimate, obtained using pilot based estimator, is improved iteratively with EM algorithm in OFDM systems. For time varying channels, pilot symbols need to be transmitted periodically to overcome the effect of fast fading. But EM-based techniques gives bandwidth advantage by using channel estimate for previous frame as the initial estimate for the next frame and avoid the need of periodic retransmission of pilot symbols. Two main points are very important in the design of a good channel estimator, namely: number of pilot symbols must be small and computational complexity should be less. In this thesis, we propose an EM-MMSE based iterative procedure to estimate the channel coefficients. The proposed technique is computationally simpler than EM technique and also yields performance improvement over pilot based techniques.