The problem of compression of correlated information across many separated encoders is well-studied in information theory, by Slepian and Wolf, and Wyner and Ziv. Their work shows that the performance of compression when there is no communication between the encoders should match the performance of compression when communication between the encoders is available. However, the early works did not have a constructive framework as to how this could be done.
Our work is the first constructive framework that shows how this can be done, in both the lossless and lossy cases. This is coined as Distributed Source Coding Using Syndromes (DISCUS), and one of the major application is in the area of distributed sensor networks. Given that the sensors are measuring the same property, and that they are correlated, how can we jointly compress the information even when the sensors do not know what the other sensors are reading? This is illustrated in the figure below:
Figure 1: Distributed Coding with no communication between the encoders
Figure 2: Illustration of a sensor network
Contact: Sandeep S. Pradhan, Julius Kusuma
The idea behind watermarking is to embed some information on a host signal in the minimally intrusive manner. It has been shown in information theory that there exists embedding methods of information at rates independent of the power of the host signal.
Recently Chen and Wornell showed that the watermarking problem can be formulated as channel coding with side information. Our construction is inspired by our discovery of the duality between source coding with side information and channel coding with side information, and takes advantage of powerful codes to achieve great performance in the face of different attacks.
Contact: Jim Chou, Sandeep S. Pradhan, Julius Kusuma