um sort of the delay a we come to my presentation on a spatial temporal signal
was
multichannel acoustic a cook
relation
uh that's
or like like to give an outline of my presentation
i would introduce my presentation by you
brightly
maybe on the problem of multichannel acoustic echo it in constellation
a a a a a a that a a approach
i i i would present
yeah
in my lecture
can say
uh
massive multichannel reproduction systems and for that i would like to
we interviews
some basic properties of a method of the channel
reproduction systems
a that i one
introduced a the novel approach for spatial can problem
preprocessing
and
give
a to or
uh um
a
a roles of this i
or practical
implementations and after that i
to
with no
so i
think
you have a
seen
similar a
block yet from
in this section so forth that
i
i is suppose you know what is says you know so we have here just a a a a a
and and and you and strong
and i i yeah
all what acoustic E
constellation it to create
a replica of sandy the or in room
um
model model as um my
F I R
filter
and the that
got
of
that my my by a filter
okay
are usually a minute
by by at least squares optimized addition
the optimal
solution
right
a or is given by the you know how to equation
also known
and a normal
equation here
we have that in terms of
are are X
it's which uh
if uh inverse of the
correlation metrics of the loudspeaker the signals in then used and
role
and are it's Y D notes
that
a cross correlation
metrics tricks between the loudspeaker signals
and the microphone signals
he's any and room
um
in
in practise
there
at the filters
are
the term and
at optimally
because
that acoustic of the own
could be
i of over at a time
is that it can be
shown
that
that's speed
of con veterans
oh adaptive filters depends of on that i gains
spread
of the autocorrelation metrics
and for some fleeting
in a typical
uh a for the the like
communication
scenarios we have only
you
if source source as in the are end room
and Z sources
are
right by
many loudspeakers speakers in there
near and room
hence we have a
and
ill conditioned
are
X six
and
how to cool
where
just problem
that i do yeah
is to preprocessed process that allows loudspeaker signal
in that near in troll in order to
decorrelate the channels with rest
but to each other
that
the preprocessing techniques
have to fulfil fill
at least
in three but
and first one is
for for of that convergence enhancement
that's
second requirement
if
is that a preprocessing technique
that doesn't introduce
all people
um uh the distortion on the signal
and
also
that
there
preprocessing technique doesn't
in use
in the yeah
uh
compute uh
uh
audition them
complexity complexity factors
for
to oh
for us
T your isn't to their production systems
you can find
many approaches
for preprocessor loudspeaker signals a
and T i mention only two
the first one was introduced by
the the is T
and to the idea
oh it is to rate up to introduce nonlinearities as the signal
this approach
feature a very low complexity
but that distortion that it
in uses
bit come quite to a problem for high quality
signals
for for them for music
conference or something like that
and uh and a second to a perceptually when motive that up holds well
introduced used by you can hear to
and the i
B is this
one have
for use on a frequency selective
phase modulation
of the a loudspeaker
signal
so the uh he of of that a loudspeaker signals
are decomposed
in in bands
in the
by means of uh analysis filter
and the sub-band signals
uh a or the face of the subband signals
and modified
in a frequency dependent
am modulating signal
and the now i
a to come to the second point
and what is like to introduce
that massive multichannel reproduction systems
the sound field
sinc to is this
systems
in at is that to use this
of the sound field to put used might have the actual source
S
with an an extended yeah yeah
V
using and i'm bill
of a speakers on the boundary of that area
here
the
the delta be in
that go
this many techniques
um you for
solving this
problem
and the
we
some of them are are this specialised
on specific
a a is you for example is that near field compensated tie i what i'm based on it
which is the
a specialised
on a circle and say and
and is
we have also for a planet and lean yeah and i is
is that
so called
it's spectral division method
it's the M
and the
more or less
is more one
uh
and i rate in an independent
technique technique it is is a a a a a W F S is the wave field synthesis techniques
which words for linear plan a and other convicts
and i image
for sure as a a low or
to kill implementations of for sound field sent to just
are
discrete
and therefore
the except
the the except if something like an aliasing frequency
and we can say that the control ability of the sound field is available only up to a given
frequency
and
here we see use this simulation one and linear very
with
twenty loudspeakers speaker
and here at the frequency have it
it's hundred
have which is below the
and you in frequency of of this yeah
right
that wave front can
be response
set correctly
and here for the two thousand
a time the test which i
all the figure that L in frequency we see that
that
where front
and a be reconstructed
firstly
so
and the
two our
problem
that we like to implement
and uh
uh acoustic echo canceller
for such a set up who with with massive multichannel
systems
this problem is
a challenging
not only be a is that is the autocorrelation matrix as
are are very last and in condition
when
addition
uh it can be shown that the someone
you to synthesized
by a massive multichannel
the production techniques
are
very sensitive to any day deviation
on the tiring
signals
uh
do you to build a preprocessing
and to hear it should be
not
that is that best section
or from from and uh so i got the point of view
that's a perception of
that is synthesized
wave fields
by
sound and to the synthesis
techniques
are
completely
i is completed different the then of stereophonic funny
the based techniques
and haynes
any
temporal preprocessing should be avoided
the so we can
do and a preprocessing
techniques
and how to cope with this problem
and
therefore we
have
it a novel approach for spatial or temporal i E C preprocessing
and this is is never a novel approach
basis
actually on how a very
all
with them
namely prevention is better than Q
and
what we would like to do
what we do
would like to do if actually
two
that
but
here and in addition and that the channel
and uh processing you
it
is that a i is that a loudspeaker signals in such a way
that is that a wave feel is set to zero euro the positions
of the i
ones
so we light
two
D lee
is are listening yeah yeah in two
two the ones
one own is quite where some microphone and i is supposed to be
and then another listening
yeah yeah
where is the desired
feel is a construct a
lee
is the
but
problem of creating
own on i i'd this not new and the in the literature
you can find many
start days
on it
and here i have listed
only some also
but to
here i i'd like to introduce
and you and a local
approach
for a
at uh the uh
creating zones of quiet with linear
and i this approach is best
it just straight
by
uh uh but is illustrated but be thing on the method of spectral division method and therefore i would like
two
just
give that say or tickle be six
of
that
mm method
and in is it is this method
assume assume what
or or to be is is actually
one the idea that some wave
feel
reproduced produced by a linear distribution
of the second outing
source of
can
B
it it's press
by a convolution of the thinks signals of
is that a second so that's was the green
function
of that services
and that that are green
functions can be and just to
as a
spatial temporal transform domain
to but also a
secondary source
in the following derivation we will restrict our configuration on a reference line
and the
uh what if we do that we can
see
is that
we have a
one time in the novel
the um
the one dimensional convolution
and for it
exact convolution theorem holds
so if we do a fourier yeah transform
we get a multiplication
and to by to be an announcement we get an expression for that mean
function
for
uh
so and uh
here is idea for creating the role of quite is actually just by multiplying
that desired wave field white i by L we don't by a specific we
and the if
we we be formally that in the wave number can mean we get them a
and expression for the desired wave field
in the wave number mean
and we get after that an expression for the subscribing
signals
that
to to prove that our approach
can i is applicable
to a a acoustic it constellation we have computed that attenuation innovations that can be done
by
this approach by creating as a of quite and we uh and
put
here
and microphone
and
right that's in an we shouldn't with that
a point which
is supposed to be
in the sweet spot
and we got an at wishing up to
seventy
D V
which is comparable was to become
yeah it cool return last
enhancement quite
S
i have a
shown
hmmm is that uh the control ability of there
wave if it's can be done only up "'cause" the anything frequent and frequency and they for for higher frequencies
acoustic it go
constellation or a typical acoustic echo equal constellation
must be still
a applied
and therefore for
we propose
here
at
divide in
the or or a yeah i don't like or decomposing "'cause" the signal into to sub bands and for a
lower
frequencies
just to so
right to use our approach and for i higher frequencies
to apply and the acoustic you constellation
yeah and say that approach
can be also
uh
use
well with the equal colour
and is
which are
more interest
interesting
for spatial audio reproduction systems
for example by applying
that output
but it represented presented by a problem
and i come to the conclusion
and the intuition of my work well
to show that is that
at to the acoustic you constellation can be done in a a distributed
manner
on the loudspeaker a
side
and on some microphone side and not
only
at it has been
the done
does on the microphones side
and i have shown one yeah
um a method to can for creating zones of quiet as linear allows loudspeaker i
and is i have sons that
that imitation that we have a
yeah is
and and on that
uh anything frequency
of the loudspeaker a rate
therefore for frequency selective implementation should be done
thank
uh we have time for questions two
T
uh i would you like
i got got a scalar
oh
i was wondering if
you have
oh
people
in side
you're or right
would they change
uh
as they move around
with you
the quiet
the G
yeah actually
we
yeah
we have
um
um
consider
only
that analytical cave yeah and we have a set for the right functions a free feel
yeah
and
hmmm
as the for the free field the green function
but for
a practical of imitation should be implemented in now
at that
man
and
so so we will have a just a mismatch between that's that's supposed
green function and a real one green function
and therefore we will not get
is that
performance of
going to quite on as we have
my my
uh what would the geometry of a relationship of the microphone
and a loudspeaker
would like thing to the lead the microphones
you're the low
result
just
in
you know maybe you
consider a circle
uh
you know where have you thought about that would be a very sad
or
a microphone geometry we we are independent actually from the microphone german trace
a a always a loudspeaker to and we have
simulated with omnidirectional
lower speed
oh
okay
and you did you in
room reflections so
so i just want
is the low frequency stuff you you do by
well making the by
yeah
it becomes an active noise can
a problem if you
microphones phones in the five
you can use that
able
to full the null
at a point to the mic
and an active noise
relation
yeah maybe
it's is a good idea i now we have a we haven't considered that
it
seems to be
with a
i
yeah think is that when you are no you the microphone
was it microphone
and the the markers
yeah
that will not
that will not that it is are yeah exactly
they were not
so
i
yeah yeah actually
therefore why have someone is that it is more interesting
to
yeah to thing to do that consideration
with in closing a right
because uh can one can include
yeah because we can define here at someone of quite
which is restricted on the position of the microphone
so only at a microphone position you
do you wouldn't
here anything
but in the listening position so for example if you can imagine
that
if
that is there a conference room and here's a table able with a microphone is on it
so it two would be with B
actually very practical
for the people to put their head on the table
the city
if they hear something on a
yeah that is actually i
a well thank you very much to little a is off
thus