Gausswindow

Gausswindow, commonly referred to as the Gaussian window, is a finite-length window function used in digital signal processing to taper data for spectral analysis and filtering. It is created by sampling a Gaussian function in time. For a window of length N, the samples are w[n] = exp(-0.5 ((n - (N-1)/2)/sigma)^2) for n = 0,...,N-1, where sigma determines the width. The window is symmetric around its center and decays toward the ends as controlled by sigma. Increasing sigma widens the window in time and narrows its main lobe in frequency; decreasing sigma produces a narrower time window and a broader frequency response.

A key property is that, in the continuous case, the Gaussian is essentially its own Fourier transform up to scaling, leading to an optimal time–frequency concentration. The discrete Gaussian window inherits this smooth, rapidly decaying sidelobe behavior, producing low spectral leakage relative to a rectangular window while avoiding abrupt edges that can cause ringing. The result is a smooth spectrum with a well-behaved trade-off between time localization and frequency resolution.

Applications include short-time Fourier transforms, spectral analysis, smoothing, and envelope estimation in audio, communications, radar, and biomedical signals. It is particularly useful when disciplined time localization is important or when a smooth taper is preferred over abruptly truncated windows. Implementation commonly involves choosing sigma relative to the window length; many numerical libraries provide Gaussian window options in their window-function collections.

Related concepts include window functions in general and the Gaussian distribution, both of which underpin the Gaussian window’s properties in time-frequency analysis.