A least-/ mean-squares approach to channel identification and equalization in OFDM

This work proposes an iterative least-/ mean-squares approach to channel identification and equalization in OFDM. This is achieved by exploiting the natural constraints imposed by the channel (sparsity and maximum delay spread) and those imposed by the transmitter (pilots, cyclic prefix, and the finite alphabet constraint). These constraints are used to reduce the number of pilots needed for channel and data recovery and also to perform this task within one packet. The diagonal nature of the OFDM channel makes it possible to perform optimal (nonlinear) mean-square detection of the data.