Joint channel estimation and data detection for OFDM systems via sphere decoding

We develop blind and semi-blind joint estimators of the channel impulse response (CIR) and data symbols for orthogonal frequency division multiplexing (OFDM) systems over a frequency selective fading channel. Using the maximum likelihood (ML) criterion, we derive two estimators for the transmit data symbols that require minimizing a complex, integer quadratic x^TGx* where x is a data vector and the matrix G characterizes each estimator. Avoiding computationally prohibitive exhaustive search, we use both sphere decoding (SD) and V-BLAST algorithms. We also modify the SD algorithm for our complex OFDM system, to handle any M-PSK, and incorporate a reduced complexity SD into the OFDM system. The quadratic for the blind estimator suffers from rank deficiency. We give an efficient solution to the rank deficiency problem. Simulation results confirm the good performance of our proposed estimators.