Year: 2009
Type: dissertation
Abstract: This thesis is concerned with relating information originating from disparate sensors by \nexploring the statistical dependence between multiple time domain signals that occur as a \nresult of a common cause influencing outputs of the sensors. Relating the signals without \nany attempt to model the environment or the behaviour of any particular object within \nit is the main focus of this work as scenarios where sufficient a-priori knowledge is not \navailable is of primary interest. Mutual Information (MI) is selected as a suitable metric \nfor determining statistical dependence mainly due to its ability to identify nonlinear high \norder e ects, and due to its ability to deal with multi-dimensional input signals with \nrelative ease. \nInspired by human perception, the focus will be on observing objects moving in space \nusing sensors that operate based on different physical principles and the fact that motion \nhas in principle, greater power to specify properties of an object than purely spatial \ninformation captured as a single observation in time. Our first intention is to utilise \nthe dependence between variables to aid active sensing. The second objective is that of \nmulti-sensor, multi-object tracking which is a challenging problem in large part because of \nthe need to solve the embedded problem of data association, which is the task of relating \nthe measurements from different sensors that correspond to the same object. \nThe contribution of this thesis include the development of a novel strategy for detecting \nthe set of signals that are statistically dependent and correspond to each other from \ntwo large multi-dimensional signal streams. The technique is based on deriving a linear \nmapping that maximises MI between the signal streams in two-dimensional space. The \nmapping is obtained by an iterative process that maximises MI through using analytical \nexpressions of the gradients of three measures equivalent to two individual entropies and \none joint entropy of the signal streams, while at the same time regularising the coefficients of the mapping using L1 and L2 norms. Thus a sparse linear mapping that makes \nit possible to identify the most mutually informative signal pairings, without the need for \nexhaustive pair-wise comparisons is obtained. This results in a common multimodal data \nassociation methodology, which could be extended to a wide range of sensors with different \nmodalities. The techniques developed are extensively analysed through a series of simulations \nand experiments. The approach is demonstrated on registration of sensors with \ndisparate modalities, registration of sensors on moving observers and target grouping.
Author: Alen Alempijevic
Institution: University of Technology Sydney
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2290683836
