| | |
Summary: ON-LINE BLIND SEPARATION OF MOVING SOUND SOURCES
J¨orn Anem¨uller and Tino Gramss ´Ýµ
Dept. of Physics, Graduate School in Psychoacoustics
Carl-von-Ossietzky-University, 26111 Oldenburg, Germany
ane@uni-oldenburg.de
ABSTRACT
In this paper, we propose a method for the on-line blind sep-
aration of sound sources in the case where the mixing filters
have a Æ-shaped impulse response. Our algorithm works en-
tirely in the frequency domain and exhibits fast convergence
due to cross-frequency couplings.
Specific problems related to on-line separation of run-
ning speech are discussed. The algorithm performs success-
ful separation of digitally mixed speech signals and of sig-
nals from a moving and a standing speaker recorded in an
anechoic chamber.
1. SOURCE SEPARATION
In the case of an instantaneous linear superposition
Ñ ´Øµ × ´Øµ (1)
of Æ independent source signals × ´Øµ resulting in Æ sen-
|
