ABSTRACT
Selection of eigentriples for efficient separation or decomposition of signals often cannot be achieved if no a priori information is available. Such information stems from diversities of the original sources or medium. This kind of information can nicely be exploited in the design of an adaptive SSA system. The behaviour of such systems is quite similar to those of conventional adaptive filters [1]. Adaptive filters are the basic structure for many adaptive and optimisation systems which broadly exist and are used for modelling natural and physical systems and plants. These systems are described by a set of parameters which can be estimated using a number of algorithms. The established learning algorithms for the estimation of adaptive filter parameters are meant for linear systems. However, there are attempts to tackle solving such problems for nonlinear systems too. Figure 4.1 shows the block diagram of a simple adaptive filter to model a linear system.
