CIF: Small: Hybrid analog-digital schemes for joint source-channel coding of digital sources

University Of Delaware, Newark DE

Investigators

Abstract

This project aims at the development of novel schemes for high throughput transmission of digital sources over noisy channels for a wide range of source and channel environments. The research effort will result in significant advancements in communications systems, including the case in which the data presents temporal or spatial correlations. Since multiple data sets, such as sensor networks and medical data, present this type of structures, this project has the potential to impact society as a whole. The main goal of this research is to study the application of hybrid analog-digital joint source-channel coding schemes for the transmission of digital sources over noisy channels. The proposed hybrid schemes consist of two sub-blocks concatenated in parallel: First, a digital-to-analog encoder, which will produce either real numbers or many multiple discrete points from the input bits. Second, a digital encoder, which will produce coded bits from the input bits. The key idea is to take advantage of the analog component of the hybrid scheme to achieve high throughput communications, and robustness against changes in the channel conditions, something that is challenging with standard digital channel codes. On the other hand, the use of pure digital-to-analog encoders would lead to significant error floors, with the consequent performance degradation. However, thanks to the use of the digital encoder sub-block, these error floors can be substantially reduced to achieve an excellent performance. Hybrid analog-digital practical coding systems have not been applied for the encoding of digital sources. The development of these techniques is of great interest in practical applications such as image and video coding, sensor networks and medical applications.