An improvement to the multiple window spectrogram using quadratic time-frequency distributions

Abstract

This paper suggests an improvement to the multiple window (MW) spectrogram. The improvement was observed when redefining the MW spectrogram in a quadratic time– frequency distribution framework. This alternate view permits the MW method to be interpreted as an application of a single ambiguity domain filter, that consists of the average of each window sequence, to the ambiguity representation of the signal. In addition, the MW spectrogram can be improved by extending the ambiguity domain filter along the Doppler lag () axis. This enhanced MW spectrogram has lower variance and bias when estimating several signals in the presence of additive white Gaussian noise, while retaining the desirable properties of the spectrogram such as realness, non–negativity and minimal cross–terms.