In turbo decoding, each component decoder must provide what are known as soft outputs for the decoded bits. To obtain these, we can use several decoding algorithms.
The BCJR algorithm is the optimal component decoder for turbo codes in terms of minimizing the decoded bit error rate. To work properly, we need to subdivide information sequence into blocks. This operation requires trellis termination of all constituent codes.
The sliding window BCJR algorithm is derived from the BCJR algorithm, permits a continuous decoding of the coded sequence without requiring trellis termination of the constituent codes and uses reduced memory span. But, number of computation is greater than that of BCJR algorithm.
In this thesis, we suggest the guideline on the determination of window width D and propose computationally efficient sliding window BCJR algorithm, in which the window moves every C times. We can improve performance and reduce complexity simultaneously with proper choice of C and window width D.
