Session S9

Session S9: Distributed Sensing and Coding

Time:                 Wednesday, May 14, 10:30-12:00
Chair:                 Lina Karam, Arizona State University
Co-Chair:          Tolga Duman, Arizona State University

S9-1: Linear Codes for Distributed Source Coding: Extracting First Order Performance From Structure
Dinesh Krithiivasan and S. Sandeep Pradhan University of Michigan, Ann Arbor MI
A new rate region is presented for a general framework of distributed source coding where the decoder is interested in lossy reconstruction of an arbitrary function of the sources. The coding scheme involves vector quantization followed by “correlated” binning. Nested linear codes are used for both quantization and binning.

S9-2: On Distributed Quantization in Scalable and Predictive Coding
Ankur Saxena and Kenneth Rose University of California, Santa Barbara CA
The design of distributed coding systems for correlated sources has been studied extensively in the literature. It is often implicitly assumed that various single source compression paradigms such as predictive coding or scalable coding may be combined with distributed coding algorithms in a straightforward manner to enjoy their respective advantages. However, this is not the case in general. In this paper we consider distributed predictive and distributed scalable coding, and show that naive combined approaches yield poor rate-distortion performance. We highlight inherent conflicts that arise between distributed quantization and predictive or scalable coding. Distributed predictive coding is further plagued by design instability of closed loop predictors, whose impact has been recognized in the context of single source coding, but is greatly exacerbated in the case of distributed coding. We propose a general framework that allows for and controls mismatch between encoder and decoder estimates in the prediction loop or the base-layer, respectively, and present iterative techniques for joint optimization of all system components. Simulation results show substantial gains over single source (separate) predictive or scalable coding techniques, as well as over naive extensions to incorporate prediction or scalability in distributed coding.

S9-3: Block-Adaptive Distributed Video Coding with Selecive Bitplane Decoding
Wei-Jung Chien, Lina Karam Arizona State University, Tempe AZ, and Glen Abousleman General Dynamics, Scottsdale AZ
This paper presents a pixel-domain block-adaptive distributed video compression (DVC) system. In the proposed system, the Wyner-Ziv frames are divided into several sub-images. Each of these sub-images is encoded and decoded independently. At the decoder, the error probability is estimated for each bitplane of the sub-images. Only the bitplanes with high error probability are Turbo decoded. A minimum-distance symbol reconstruction is proposed to estimate the rest of the bitplanes by using the side information and Turbo decoded bitplanes. More accurate placement of the parity bits results in an improved system performance, especially for video sequences with a relatively large static background. Coding results and comparison with existing DVC schemes and with H.264 intraframe coding are presented to illustrate the performance of the proposed system.

S9-4: Analysis of Distributed Estimation over Fading Multiple Access Channels with Channel State Information at the Sensors
Mahesh Banavar, Cihan Tepedelenlioglu and Andreas Spanias Arizona State University, Tempe AZ
We consider a wireless sensor network for estimating a random parameter. The sensors transmit their observations via fading channels to a fusion center, where the parameter is estimated. We calculate the variance of the estimate when we have full, partial and no channel information at the sensors. We compare these with a benchmark obtained when the channels are additive white Gaussian noise. We show that for a wide range of scenarios, partial CSI at the sensor is sufficient to get reliable estimates of the parameter.

S9-5: Sparse Representation for Pattern Classification using Learned Dictionaries
Jayaraman J. Thiagarajan, Karthikeyan N. Ramamurthy and Andreas Spanias Arizona State University, Tempe AZ
Sparse representations have been often used for inverse problems in image processing. Furthermore, frameworks for signal classification using over-complete representations have also been developed. A number of dictionary learning algorithms have been proposed to generate data-driven sparse representations. These iteratively update the over-complete dictionary from which the sparse representation is obtained. K-SVD is an existing dictionary learning algorithm that has been very effective for image denoising applications. In this research, we provide a framework for template based pattern classification using sparse and over-complete representation over learned dictionaries.