Abstract:
Considering the high cost and complexity of multiple-channel blind source separation, single-channel solutions have elicited a significant amount of interest in source se...Show MoreMetadata
Abstract:
Considering the high cost and complexity of multiple-channel blind source separation, single-channel solutions have elicited a significant amount of interest in source separation. Joint maximum likelihood sequence estimation implemented by the Viterbi algorithm (VA) is a seemingly ultimate solution to the problem of data transmission over dispersive channels. However, both joint reduced-state sequence estimation (JRSSE) and joint delayed-decision feedback sequence estimation (JDDFSE) have numerous shortcomings when applied to single-channel blind signal separation. Hence, joint double-direction DDFSE (JDD-DDFSE) is proposed in this article. JDD-DDFSE employs joint VA (JVA) of the low-order channel twice. The estimated symbols from the initial JVA are utilized to equalize the causal and non-causal taps of the channel beyond the trellis state in the second JVA. A method that can derive the log likelihood ratio of the estimated symbols is provided; the complexity of this method is considerably lower than that of soft-output VA or the BCJR algorithm. Hence, the entire computation complexity of the iterative separation structure basically remains unchanged compared with that of the no-iteration structure. For the non-causal channel, the proposed algorithm has the lowest computation complexity and demonstrates a separation performance similar to that of conventional JVA, JRSSE, and JDDFSE algorithms.
Published in: IEEE Transactions on Vehicular Technology ( Volume: 69, Issue: 2, February 2020)