In order to analyse and interpret speech signals, different time-frequ
ency representations are used (e.g. spectrogram, Wigner-Ville distribu
tion, wavelets). In this paper we construct within Cohen's class of ti
me-frequency distributions the distribution that is optimally suited f
or the representation of speech signals. Thereby we take advantage of
the special time-frequency structure of speech expressed in the Elemen
tary Waveform Speech Model (EWSM, d'Alessandro, 1990). As an applicati
on we present an algorithm that extracts a point pattern in the time-f
requency plane out of the speech signal using the optimized distributi
on. Thus we get a very simple representation of the speech signal that
is well interpretable both for non-stationary and for stationary spee
ch segments. Furthermore this represention could serve as a base for f
urther analysis (e.g. classification).