Multiple Model Framework for Noncoherent Detection in Communications

点击翻译按钮获取中文摘要

A fundamental problem in communications is that of detecting an information-carrying signal which has been transmitted over a channel subject to interference, noise and other impairments. This project studies this problem, with a special focus on the case when the structure of the transmission channel and the noise/interference is unknown to the receiver. One typical instance of this situation is suppression of interference that consists of signals from an unknown number of other transmitters who utilize the same frequency band. This is relevant to several practical problems: multiuser detection for the third generation spread spectrum (CDMA) system, and antijamming for the Global Positioning System (GPS) or wireless communications systems, for example. In summary, the outcome of the research will help improve the quality of wireless links and enable a more efficient use of the radio spectrum. The project takes a systematic, decision-theoretic approach to the problem under study. A new paradigm is developed, where the key idea is to use mixture models (a.k.a. ``mixture of experts'') for the received data. Specific topics addressed include: 1) Formulation of a general model/framework for communications signals via mixture densities. 2) Development of statistically sound methods for the mixture model parameter estimation. 3) Design of demodulation algorithms that utilize the mixture model formulation of the received data, and that can be integrated with soft-decision decodable GF(2) codes. 4) Algorithm design for practical problems including: suppression of interference with unknown structure; methods that can trade off between ``noise-prewhitening'' interference suppression and multiuser (joint) detection; diversity combination with an unknown number of ``significant'' paths (e.g., for CDMA or UWB); demodulation of signals with unknown modulation formats; suppression of non-Gaussian noise.