Multi-Channel
Signal
Separation
Based
on
Decorrelation
RLE
Technical
Report
No.
573
Ehud
Weinstein,
Meir
Feder
and
Alan
V.
Oppenheim
March
1992
Research
Laboratory
of
Electronics
Massachusetts
Institute
of
Technology
Cambridge,
MA
02139
USA
Multi-Channel
Signal
Separation
Based
on
Decorrelation
RLE
Technical
Report
No.
573
Ehud
Weinstein,
Meir
Feder,
and
Alan
V.
Oppenheim
March
1992
Research
Laboratory
of
Electronics
Massachusetts
Institute
of
Technology
Cambridge,
MA
02139
USA
This
work
was
supported
in
part
by
the
U.S.
Air
Force
Office
of
Scientific
Research
under
Grant
AFOSR-91-0034,
in
part
by
the
U.S.
Navy
Office
of
Naval
Research
under
Grant
N00014-90-J-1109,
in
part
by
the
Defense
Advanced
Research
Projects
Agency
monitored
by
the
Office
of
Naval
Research
under
Grant
N00014-89-J-1489,
and
in
part
by
the
Wolfson
Research
Awards
administered
by
the
Israel
Academy
of
Sciences
and
Humanities.
Multi-Channel
Signal
Separation
Based
on
Decorrelation
Ehud
Weinstein'
Meir
Feder
2
Alan
V.
Oppenheim
3
Abstract
In
a
variety of
contexts, observations
are
made of
the
outputs
of an unknown
multiple-input
multiple-output
linear system,
from
which
it
is
of
interest
to
recover
the
input
signals.
For
example,
in
problems of
enhancing
speech
in
the
presence
of
background
noise, or
separating
competing
speakers,
multiple
microphone
measurements
will
typically
have
components
from
both
sources,
with
the
linear
system representing
the
effect
of
the
acoustic
environment. In
this
paper
we
consider
specifically
the
two-channel
case
in
which
we
observe
the
outputs
of
a
2
x
2
linear
time invariant
system
with
inputs
being
sample functions
from
mutually
uncorrelated
stochastic
processes.
Our
approach
consists
of
reconstructing
the
input
signals
by
making
an
essential
use
of
the
assumption
that
they
are
statistically
uncorrelated.
As
a special
case,
the
proposed approach
suggests
a
potentially
interesting
modification of
Widrow's
least
squares
method
for noise
cancellation,
when
the
reference
signal contains
a
component
of
the
desired
signal.
'Ehud
Weinstein
is
with
the
Department
of
Electrical Engineering
-
Systems,
Faculty
of
Engineering,
Tel-Aviv
University,
Tel-Aviv,
69978,
Israel.
He
is
also
an
adjunct
scientist
at
the
Department
of
Applied
Ocean Physics
and
Engineering,
Woods
Hole
Oceanographic
Institution.
2
Meir
Feder
is
with
the
Department
of
Electrical
Engineering
-
Systems,
Faculty of
Engineering,
Tel-Aviv
Univer-
sity,
Tel-Aviv,
69978,
Israel.
3
Alan
V.
Oppenheim
is
with
the Department
of
Electrical Engineering and
Computer
Sciences,
Massachusetts
Institute
of
Technology
Cambridge,
MA.
02139,
USA.