An Algorithmic Study of Optimal Multiresolution Quantization and Joint Source-Channel Coding

Polytechnic University Of New York, Brooklyn NY

Investigators

Abstract

ABSTRACT 0208678 Xiaolin Wu and Nasir Memon Polytechnic Univ of NY An Algorithmic Study of Optimal Multiresolution Quantization and Joint Source-Channel Coding In this research project the investigators will develop algorithms to improve service quality of video and audio streaming over switched networks, wired or wireless, in adverse network conditions such as congestion, delay, and packet loss. The idea is to have graceful degradation in quality of service rather than outright stoppage of audio and video playback, when the effective transmission rate drops. This will significantly enrich users' experience with IP-based multimedia services. The research consists of two inter-connected studies: optimal multiresolution (progressively refinable) quantization (MRQ), and uneven error-protected packetization (UEPP) of scalable source code streams generated by MRQ. The investigators' main objective is to develop efficient algorithms for optimal MRQ design, and for optimal bit allocation between scalable source codes of MRQ and forward error correction codes. Their approach differs from the existing rich body of literature on joint source-channel coding and on MRQ in that the problem is treated as one of combinatorial optimization, rather than resorting to continuous Lagrangian optimization. The investigators will classify various optimization problems related to MRQ design and joint source-channel coding by their innate computational tractability, and develop either globally optimal algorithms for those that are solvable in polynomial time, or heuristic algorithms for those that are NP-hard. In order to obtain efficient algorithms for MRQ and UEPP, discrete structures of the underlying objective functions of optimization will be explored, which in turn may offer new insights into these problems of increasing interest and immediate relevance to internet and wireless communications.