Interpolation by Convolutional Codes, Overload Distortion, and the Erasure Channel

This paper is motivated by sequence based methods of quantization, specifically trellis coded quantization as described by Marcellin and Fischer. In this method n bits specify one of 2 sup n different quantizers, and consecutive outputs of a rate k/n convolutional code specify admissible sequences of quantizers. We investigate how closely randomly generated binary source sequences can be matched by codewords from a convolutional code What distinguishes this paper from prior work is that a randomly selected subsequence with density lambda is to be approximated as closely as possible