Evaluation of the optimal lengths and number of multiple windows for spectrogram estimation of SSVEP

Abstract 

The purpose of this paper is to present the optimal number of windows and window lengths using multiple window spectrogram for estimation of non-stationary processes with shorter or longer duration. Such processes could start in the EEG as a result of a stimuli, e.g., steady-state visual evoked potentials (SSVEP). In many applications, the Welch method is used with standard set-ups for window lengths and number of averaged spectra/spectrograms. This paper optimizes the window lengths and number of windows of the Welch method and other more recent, socalled, multiple window or multitaper methods and compares the mean squared errors of these methods. Approximative formulas for the choice of optimal number of windows and window lengths are also given. Examples of spectrogram estimation of SSVEP are shown.

Keywords: Time–frequency analysis, Multitaper, Multiple window spectrogram, EEG, Steady-state visual evoked potentials, SSVEP, Welch method, WOSA, Peak matched multiple windows, Locally stationary processes, Hermite functions