Multitask Sparse Bayesian Channel Estimation for Turbo Equalization in Underwater Acoustic Communications
Yaokun Liang, Hua Yu, Fei Ji, Fangjiong Chen
Abstract
In this article, a multitask sparse Bayesian learning channel estimation based on factor graphs (MT-SBL-FG) for turbo equalization (TEQ) in underwater acoustic (UWA) communications is proposed. Based on the framework of multitask sparse Bayesian learning, the received data are divided into subblocks and the channel estimation (CE) of each subblock is regarded as a task. Considering the time-varying channel in UWA communications and the CE error in turbo equalization, a factor graph is proposed for the multitask CE of the subblocks. Based on the factor graph, a message-passing algorithm with weighting factor is derived. The performance of channel estimation is improved by utilizing the temporal correlation between the subblocks and the sparsity of the UWA channel. The proposed algorithm has been tested by the underwater trial data collected in an experiment conducted in Qiandao Lake, Hangzhou, China, in May 2016. We have verified that the proposed algorithm outperforms other CE algorithms in terms of the bit error rate (BER).