Turbo Decoding Complexity Reduction by Symbol Selection and Partial Iterations

Based on an analysis on the recursive computation of the iterative maximum a posteriori (MAP) algorithm for turbo decoding, this paper considers a modified MAP scheme with reduced block lengths for symbols with unreliable detection after some initial iterations. Applying symbol selection based on cross-entropy measurement for parallel concatenated convolutional codes, we develop partial, windowed iterations for selected symbols. By omitting computations for symbols with reliable detection results, this approach significantly reduces complexity but well maintains the performance by complete iterations.