Unequal error protection of subband coded bits

Abstract

Source coded data can be separated into different classes based on their susceptibility to channel errors. Errors in the Important bits cause greater distortion in the reconstructed signal. This thesis presents an Unequal Error Protection scheme for selectively protecting subband coded bits. A subband coder is designed to provide encoded bits with different levels of importance. An Autoregressive source of the first order is used as the source model. Unequal error protection is provided by a bank of channel coders, each with a different level of error protection for the different classes of output from the subband coder. Each of the channel coders is simulated as a coded modulation scheme with an underlying convolutional code to map coded bits into 4-PSK symbols for transmission over an AWGN channel. The individual performances of the subband coder and the channel coder are presented. Subband coding is shown to be an useful method for separating input into bit streams of varying levels of importance. The performance of an overall system consisting of the subband coder and the channel coder is studied. The distortion in the reconstructed signal at the receiver was more when there were errors in the important bits. Errors in the less important bits caused a very small increase in the overall distortion. This implied that the important bits should be given a preferential treatment by allotting as much of available channel redundancy to them as possible.