ITR: Error Resilient Communication and Adaptive Data Compression over Wireless and Noisy Channels

Brandeis University, Waltham MA

Investigators

Abstract

With many communications applications, such as mobile computing, as well as in data storage, data compression is essential for large sets of digital information such as images, video, and multi-media. Adaptive methods that can in real time "learn" about the data to compress it well are the most powerful, but can have the drawback that a single error can propagate and corrupt all data to follow. Based on past work by the PI, this project will study error resilient communication protocols and adaptive real-time data compression where error propagation is essentially prevented. This new work is targeted at both lossless and lossy applications, including image and video compression. The project will consider how the proposed techniques can be implemented in a way consistent with existing video compression standards. It will study how these techniques can be combined with reversible variable length codes (used in the MPEG-4 standard) to improve recovery from a catastrophic error burst. Although reversible codes have been studied by a number of authors in the past, there is much that is not known about efficient optimal methods. The project will also investigate how the learning employed by error resilient adaptive compression methods can provide a filtering mechanism that can be used for fast browsing of large data over a noisy channel, a situation that is increasingly important. The work will include theoretical analysis, algorithms design, and experimental work.