Diversity systems for Rayleigh fading channels: an application of multiple description source codes

Abstract

We consider the transmission of a memoryless Gaussian source over Rayleigh slow-fading channels. A signal transmitted through such channels is not only disturbed by Gaussian noise but also suffers random amplitude attenuation. The classical approach to overcome fading is using diversity techniques. In conventional diversity techniques, the same information is sent over L > 2 channels. The error probability is reduced by properly combining the L received signals, e.g., equal- gain combining or maximum ratio combining (MRC). These techniques have been well established and widely used. However nowadays the communication systems are required to transmit more and more information. Instead of transmitting identical information, we must consider transmitting different information over each channel and still maintaining low probability of errors. Channel coding has been shown to be an effective approach to accomplish this job. But in order to maintain low probability of errors, the channel code must be interleaved before transmission. Thus, channel coding provides a better approach than conventional diversity techniques for overcoming fading but pays the price of longer interleaving delay. In two-way communication systems, the interleaving delay is an important issue. In this work, we introduce a combined source and channel coding approach, the multiple description source code based system. The multiple description source code is designed to trans ml it different information over each channel and if all the information are received correctly, a higher fidelity can be achieved at the receiver, but if part of the information gets lost, a minimum fidelity is still guaranteed. We show that the multiple description scalar quantizer (MDSQ)-based system achieves better performance than the MRC-based system without introducing additional interleaving delay, and based on the same interleaving delay it also outperforms the channel coding based system. We also show that the MDSQ-based system can have different number of quantization levels at fixed rates therefore can have different performance behavior at fixed channel SNR'S. Due to this flexibility, the MDSQ makes the job of designing a dual diversity communication system become much easier.