In many communications systems, data can be divided into different importance levels. For these systems, unequal error protection (UEP) techniques are used to guarantee lower BER for the more important classes. In particular, if the precise characteristics of the channel are not known, UEP can be used to recover the more important classes even in poor receiving conditions. In this paper, we derive bounds on the performance of unequal error protecting turbo codes. These bounds serve as an important tool in predicting the performance of these codes. In order to derive the bounds, we introduce the notion of UEP uniform interleaver which is a random interleaver that does not change the order of classes in the turbo code frame. We also present a method to derive the weight enumerating function for UEP turbo codes.

A major concern in designing communication systems is to maintain quality of service for a wide range of channel conditions. This is an important issue particularly for the applications where precise characterization of the channel is impossible. For such applications, the source data can be classified into several classes and Unequal Error Protection (UEP) can be used to effectively protect the more important classes even in poor receiving conditions. This paper is focused on the study, design, and performance evaluation of unequal error protecting turbo coded modulation schemes. We first propose several schemes for unequal error protecting using turbo coded modulation. All these schemes provide high performance gains for more important classes that can hardly be achieved using conventional coded modulation schemes. We then study unequal error protecting turbo coded modulation schemes by deriving channel capacity and cutoff rates for different protection levels. We show that for more important classes more room is available for improvement.

In this paper, performance of turbo trellis coded modulation (turbo-TCM) is investigated in the Gilbert-Elliot channel model. We consider three methods for transmission through this channel. In the first method no channel interleaver is used, in the second method we employ a channel symbol interleaver after the mapping to make the channel memoryless. Finally, in the last method a bit interleaver is used before mapping to increase to diversity. For all cases, we have assumed that the channel state is known at the receiver.