Scalar quantization of memoryless sources over memoryless channels using rate-one convolutional codes

Source and channel coding techniques are widely used in many communication systems. In source coding, the objective is to retain the essential information of the source by removing its redundancy. Channel coding on the other hand, introduces redundancy in order to provide protection against channel noise. When designing channel codes, it is common to assume that the source encoder has removed all of the source redundancy. In some practical situations, however, the source encoder fails to entirely eliminate the source redundancy. In these cases, the redundancy remaining at the output of the source encoder - the 'residual redundancy' - may be used to combat channel noise. In this paper, the case where the residual redundancy is strictly in the form of a non-uniform distribution is considered. Instead of transmitting the quantizer output directly over the channel, it is first encoded by a rate-one convolutional encoder. The purpose of the convolutional encoder is not to introduce additional redundancy but to convert the residual redundancy into a more usable form.