Rank-adaptive signal processing (RASP) a subspace approach to biological signal analysis .II. Applications

In many biomedical signal processing problems, the signal of interest is corrupted by noise and interference from other sources. The nonlinear and time-varying nature of biological systems, inter- and intra-patient variability, constraints imposed by patient safety and the desire for less or noninvasive monitoring, exemplify the problems that must be overcome by signal processing. We present experimental results that illustrate the usefulness of the subspace approach in a variety of practical applications. In contrast to linear time-invariant (LTI) and conventional adaptive filters, the subspace approach requires no reference input or a priori knowledge of the frequency contents of the data. The signal and noise subspaces are determined directly from the gaps in the singular value spectrum.