Characterisation of signal in terms of their nonlinear nature has emerged in physics in the mid 1990s and is only now being adopted in signal processing. This talk will highlight the need to assess the nonlinearity within a signal prior to choosing the actual processing models, since e.g. the change from the stochastic to deterministic behaviour in heart rhythm might indicate heart attack. Next, the link between signal nonlinearity characterisation and the modelling beyond second order statistics will be elucidated, and the intricate relationship between the local predictability in phase space and the degree of signal nonlinearity will be analysed. In the next step, the Delay Vector Variance (DVV) method for the simulatenous assessment of signal nonlinearity and uncertainty will be introduced, and several case studies, including fMRI, sleep psychology, and qualitative assessment of machine learning algorithms will be presented. Finally, some emerging online approaches will complete the talk.
Cite as: Mandic, D. (2007) Exploiting nonlinearity in signal processing: qualitative assessment of adaptive filtering algorithms and signal modality characterisation. Proc. ITRW on Nonlinear Speech Processing (NOLISP 2007)
@inproceedings{mandic07_nolisp,
author={Danilo Mandic},
title={{Exploiting nonlinearity in signal processing: qualitative assessment of adaptive filtering algorithms and signal modality characterisation}},
year=2007,
booktitle={Proc. ITRW on Nonlinear Speech Processing (NOLISP 2007)}
}