Neural networks based parallel Viterbi decoder by hybrid design

A hybrid scheme integrating analog and digital methods is presented to design a Viterbi decoder based on neural networks. Due to its fully parallel architecture, neural networks based Viterbi decoder is significantly faster than the purely digital decoder. The fully parallel structure is obtained by implementing the branch metric calculation and add-compare-select (ACS) using the neural networks while the register exchange using parallel digital circuits. The hybrid Viterbi decoder is more suitable for VLSI implementation.