TABLE OF CONTENTS An iterative turbo decoder consists of two component decoders concatenated serially via an interleaver, identical to the one in the encoder. SISO ( Soft Input/Soft Output) algorithms are well suited for iterative decoding because they accept a priori information at their input and produce a posteriori information at their output. In turbo decoding, trellis based decoding algorithms are used. These are recursive methods suitable for the estimation of the state sequence of a discrete-time finite-state Markov process observed in memoryless noise. With reference to decoding of noisy coded sequences, the MAP algorithm is used to estimate the most likely information bit to have been transmitted in a coded sequence. Here, we only discuss the iterative decoding of two-dimensional turbo codes. The extension to the case of multidimensional concatenated codes is straightforward.
The Bahl, Cocke, Jelinek, and Raviv (BCJR) algorithm [57], also known as the forward-backward or the a posteriori probability algorithm, or Maximum a posteriori algorithm, is the core component in many iterative detection and decoding schemes. BCJR algorithm is optimal for estimating the states or the outputs of a Markov process observed in white noise. It produces the sequence of A Posteriori Probabilities (APP), P( u k= i observation), i = 0, 1, where P( u k = i observation) is the APP of the data bit u k given all the received sequence. The...
