Introduction to Time-Frequency Signal Analysis

A nonparametric feature for neonatal EEG seizure detection based on a representation of pseudo-periodicity

“…The use of the product of two functions in the definition means that there can be a large number of choices of a i (t) and f i (t) that result in seizure(t). A practically useful definition of a i (t) and f i (t) can be obtained by ensuring the frequency content of the AM is significantly less than the IF [23,24].…”

“…As the interference tends to be highly oscillatory a simple method of reducing it, is to filter or smooth the WVD with a 2D lowpass filter [22]. The smoothness of a TFD can be improved by increasing the effective bandwidth duration (BT) product of the smoothing filter [23].…”

The T-class of time–frequency distributions: Time-only kernels with amplitude estimation

“…Joint time -frequency representations (a.k.a time -frequency distributions, TFDs) are transformations that describe the energy density of the signal simultaneously in time and frequency. Most important TFDs are members of the quadratic (bilinear) class [1]- [4]. The early examples of the quadratic class include the spectrogram and the Wigner-Ville distribution (WVD).…”

“…This class has originated from the work of Wigner in 1932 [5], formulated with the analytic signal by Ville in 1948 [6], then generalized by the work of Cohen in quantum physics in 1966 [7], hence it is also known as Cohen's Class. However, quadratic TFDs of multicomponent signals suffer from the presence of cross-terms [1]- [11], which can obscure the real features of interest in the signal. The time -frequency resolution (energy concentration) is an important issue for both mono-and multicomponent signals, especially for non-parametric IF estimation [1].…”

Multiple-view time–frequency distribution based on the empirical mode decomposition

“…where, s k (t) = a k (t) cos (φ k (t)) is a signal component or monocomponent, M is the number of components, a k (t) is the amplitude and φ k (t) is the phase (which is the anti-derivative of the FM or instantaneous frequency (IF)) of the k th component, [5]. The use of (1) as a nonstationary signal model results in signal components that are not unique.…”