ABSTRACT
A major problem in the analysis of biological signals is the extraction of useful information, such as significant frequency content, from brief noisy sequences which may be unequally spaced or missing some data. This chapter describes and illustrates two methods, fast orthogonal search for spectral analysis, and parallel cascade identification for modeling dynamic nonlinear systems and individual sequences of time-series data. It shows that fast orthogonal search is effective with brief, noisy signals and with unequally spaced data, and can achieve eight times the resolution of a Fourier series analysis. The chapter briefly describes sinusoidal analysis via fast orthogonal search, and system and signal modeling via parallel cascades. It illustrates both methods on synthetic data and parallel cascade on data obtained from the human visual tracking system. The chapter illustrates how the parallel cascade method can be used to analyze a single sequence of time-series data.
