A unified approach to the Viterbi algorithm state metric update for shift register processes

A unified approach to the scheduling of the Viterbi algorithm state metric update is proposed for single and multiple input shift register processes of arbitrary radix. Using a systematic recursive procedure, the trellis iteration is decomposed into smaller independent subtrellises. The decomposition is organized using a matrix framework, from which in-place and constant geometry schedules can be generated. Iteration of the trellis corresponds to iteration of the subtrellis within each column of a state metric matrix. This leads to the classification of architectures as column sequential or column concurrent, depending on the assignment of add-compare-select processors to columns. Based on different initial partitions of the state metric matrix a family of architectures is derived, including a new class of parallel cascaded Viterbi decoders, which extend the throughput of existing cascaded decoders