We proposed an algorithm which
enables to detect the number of
speakers, accurate s and spectral envelopes from co-channel input
simultaneous speech signals with spectral domain procedure.
It showed a high performance for speech signals of both single speaker
and two speakers.
Still, several improvements are prospective by
considering temporal continuity of
contour (e.g., introducing
Fujisaki model), incorporating variance
into the model parameters also as a variable or by introducing a priori
probability distribution of the model parameters, etc.