Viterbi Equalization Using Nearest Constellation State Reduction

The patent segments the decoding process into two distinct stages: a symbol-reduction stage that performs preliminary processing to eliminate improbable symbols, and a Viterbi decoding stage that processes only the reduced set of probable symbols. This segmentation divides the originally monolithic high-complexity decoding task into manageable parts, where the symbol-reduction device handles the bulk of the filtering work before the Viterbi decoder processes the refined symbol set.

The symbol-reduction device performs preliminary action by determining additional information about possible consequential states before the Viterbi decoding begins. This preliminary processing eliminates improbable symbols in advance, so that when the Viterbi decoder operates, it only needs to process a significantly reduced set of candidate symbols, thereby reducing the steep processing cost associated with large modulation stages and long code memory.

The patent merges the advantages of both MMSE-DFE and Viterbi decoders by combining the symbol-reduction capability of MMSE-DFE with the optimal path-search capability of Viterbi decoding. The symbol-reduction device uses MMSE-DFE-like processing to reduce symbol candidates, while the Viterbi decoder maintains detection security by performing optimal path selection on the reduced set, achieving both low processing cost and high detection security.

The patent applies local quality by using different processing strategies for different parts of the decoding process: the symbol-reduction device uses a simplified MMSE-DFE-like approach for preliminary filtering, while the Viterbi decoder applies rigorous optimal path-search for final decision-making. This localized application of different quality levels ensures computational efficiency in the reduction phase while maintaining high detection security in the final decoding phase.