Universal sequential decoding

An algorithm is proposed for universal decoding of convolutional/trellis codes employed over unknown channels. On discrete memoryless channels and at rates below the computational cut-off rate (corresponding to a uniform input distribution), the algorithm achieves an asymptotic complexity-vs.-performance tradeoff similar to the tradeoff achieved by the Viterbi (1979) algorithm, but with the benefit that the algorithm's implementation does not require knowledge of the channel law. The algorithm is also applicable to finite-state channels and to other families.