Summary: Correlated modulation: a criterion for blind source separation
Medical Physics Group and Graduate School in Psychoacoustics
Carl von OssietzkyUniversity, D26111 Oldenburg, Germany
Summary: The problem of blindly separating a convolutive mixture of modulated signals is consid
ered. Spectrograms of the signals are computed and separation is performed in the frequency domain.
A new algorithm for blind source separation is proposed, which is based on correlated modulation in
the sources' different frequency channels. For example, speech contains correlated modulation in dif
ferent frequency regions. The algorithm successfully separates mixtures of modulated artificial signals
and of speech.
To appear in: Joint meeting of the Acoustical Society of America and the European Acoustics Association,
Berlin, Germany, March 1419, 1999.
The goal of blind source separation is to reconstruct mutually independent source signals s 1 (t); : : : ;
s N (t) when only mixtures m 1 (t); : : : ; mM (t) of them can be observed. A minimum set of assump
tions regarding the probability density functions or the autocorrelation functions of the sources are
made in order to accomplish this. In particular, the geometry of the sensors and sources is com
pletely unknown, hence the term `blind'.
The case of an instantaneous, linear, square and invertible mixing of the sources is wellunderstood.