Sequential Decoding for Mobile Receivers

Currently many classical detection and decoding algorithms receive new attention in mobile communications. Amongst those classics are low density parity check codes [1], concatenated codes [2], weighted decoding with log-likelihood values as well as soft-in/soft-out decoding as performed by the BCJR algorithm and its derivatives [4], [5]. All four classics show up as key elements in the so-called turbo decoding which first was invented by Berrou et al. in [6] as an iterative scheme to decode parallel concatenated codes. Later it was recognized that this scheme, namely iterative decoding by exchanging extrinsic values between two or more decoders, is a general principle named the ”Turbo Principle” [5]. In this talk we will revive another classical decoding method called sequential decoding in order to use it for iterative decoding in high memory situations. For iterative detection/decoding (turbo) schemes in mobile receivers we modify sequential decoding which contrary to APP (BCJR) decoding enjoys a complexity almost independent from the number of states. This novel LISt-Sequential (LISS) decoder avoids most of the drawbacks of the classical sequential decoders such as variable work load and erased frames when working within a turbo scheme. It uses are a metric containing a priori and channel values, a metric length bias term for speeding up the tree search, a soft extension of paths without increasing the stack size and soft weighting to obtain a soft output. We present several mobile turbo applications using the LISS including equalization, single antenna interference cancellation, multiuser and MIMO detection. It is shown [7] that the LISS achieves the optimal (genie) or APP performance at a BER after the outer decoder of less than 10−4, but it works also at channels with a high number of taps, constellation points, antennas and states where APP (BCJR) algorithm become infeasible.We have presented the list-sequential algorithm (LISS) which is a tree based soft-in/soft-out algorithm to be used when trellis based algorithms fail due to a excessively high number of states. It is controlled by a single parameter the size of the stack memory supplied. It works only in connection with a turbo scheme because it relies on a priori information to improve its tree search. The search is controlled by a priori-, channeland length bias term metrics. The soft features are extensively used, namely as soft channel inputs including the channel state (fading) information of mobile receivers, soft a priori values, soft augmentation of paths and soft weighting to generate a soft output. The LISS is easily extended to a bidirectional LISS which uses the same stack and does not require merging of paths as in the classical bidirectional sequential decoders [7]. We discuss a variety of applications but many more are conceivable and are currently studied such as high memory convolutional codes, joint channel estimation in varying mobile channels, as well as detection with a LISS and joint source and channel coding. In comparison to other algorithms the main question of the new receiver algorithm is complexity. Note, that the only parameter which controls the complexity and the performance of the LISS algorithm is the size of the stack memory. Therefore high and low end implementations of the LISS can be offered which is a nice feature most welcomed in mobile receivers .