Second European Conference on Speech Communication and Technology

Genova, Italy
September 24-26, 1991


An Auditorily Based Spectral Transformation of Speech Signals

Manfred Beham

Lehrstuhl für Datenverarbeitung, Technical University of Munich

This paper describes a special perceptually based spectral transformation which takes into account basic properties of the auditory system. The main feature of this spectral transformation which has been especially designed as preprocessing stage for an automatic speech recognition system is the possibility of simple recursive calculation. Aurally adequate spectrum analysis is performed by a modified Fourier-t-Transformation which allows the adjustment of time and frequency resolution to the frequency dependency of the human ear. Psychoacoustic data is the basis for these resolution capabilities. Keywords: Speech Signal Preprocessing, Auditorily Based Spectral Transformation, Fourier-t-Transformation, auditory model

Full Paper

Bibliographic reference.  Beham, Manfred (1991): "An auditorily based spectral transformation of speech signals", In EUROSPEECH-1991, 1437-1440.