Matched Gaussian multitaper spectrogram

A novel multitaper spectrogram estimator for Gaussian functions is proposed. The multitapers are the Hermite functions and a fixed number of few multitapers are used in the estimate. The weighting factors of the different spectrogram functions are optimized to give the approximative Wigner distribution for the Gaussian function. The performance of the estimator is investigated in terms of resolution and cross-term reduction in the time-frequency domain. Additionally, a simulation example shows the robustness against white noise disturbances. The performance of the new estimator is compared to the Wigner distribution, the usual spectrogram as well as the Choi-Williams and the Born-Jordan distributions.