Joint data and channel estimation using fast blind trellis search techniques

Proposes a novel method for blind sequence estimation of data that are transmitted over unknown linear distortive channels. For every possible sequence that can be transmitted, the procedure finds the best possible channel fit corresponding to the noisy channel output sequence (typically using the least squares procedure). The data and the channel are estimated to be those with the overall best fit. This simple but exhaustive search procedure for blind channel equalization is noninstrumentable because of the exponential growth in its complexity with the length of the data sequence. The author proposes a suboptimal trellis search algorithm that strives to achieve the optimal performance with only a linear complexity. Fast convergence of the algorithm in estimating the channel is demonstrated for binary pulse amplitude modulation (2-PAM) and a variety of channels. Convergence at high signal-to-noise ratios (30 dB) typically occurs within 50 symbols for the channels considered. At a low SNR (10 dB), convergence occurs within 100 symbols. The number of states in the decoder trellis, which is a measure of the decoding complexity, is M^L, where M is the number of modulation levels and L is the channel memory