Newbie Understanding Beamforming and Steering

Hi, I'm not sure whether to be posting a new topic or piggybacking on the
backs of many others, I've got so many questions  If people can point me
at some literature that can point me in the right direction or have some
comments about my proposed plan I'd be very grateful.

For my final year project I want to take a crude look at signal source
acquisition.  My plan was to take some recordings using two cardiod
microphones and be able to 'listen' in certain directions, such as towards
one of two people speaking.  I understand I can apply weighted delays to
the incoming sounds to attenuate the surrounding signals from the one I'm
trying to listen to.  In my case is it as simple as having two circular
buffers holding the incoming audio data and just applying a delay to one
side or the other, then once a certain level of confidence has been found
between the similarities of the two signals I can assume the beam is being
steered in the direction I require?  Once a certain level of confidence has
been achieved is it as simple as summing the two signals together to create
a single mono channel with the required audio in it?

Compared to many of the current questions being asked this is very
trivial.

Many thanks in advance, apologies if this seems like some daft questions,
Andrew

flamingo11 skrev:
> Hi, I'm not sure whether to be posting a new topic or piggybacking on the
> backs of many others, I've got so many questions  If people can point me
> at some literature that can point me in the right direction or have some
> comments about my proposed plan I'd be very grateful.

> For my final year project I want to take a crude look at signal source
> acquisition.

Hmm... exactly what do you mean by "source acquisition"?
Most people are familiar with "signal acquisition"...

> My plan was to take some recordings using two cardiod
> microphones and be able to 'listen' in certain directions, such as towards
> one of two people speaking.  I understand I can apply weighted delays to
> the incoming sounds to attenuate the surrounding signals from the one I'm
> trying to listen to.

In principle, this is correct, yes.

> In my case is it as simple as having two circular
> buffers holding the incoming audio data and just applying a delay to one
> side or the other, then once a certain level of confidence has been found
> between the similarities of the two signals I can assume the beam is being
> steered in the direction I require?

No. You could have locked on to that noisy AC unit humming
in the corner of the room.

> Once a certain level of confidence has
> been achieved is it as simple as summing the two signals together to create
> a single mono channel with the required audio in it?

Well... yes and no. The operations required to achieve
beamforming are as simple as summing weighted signals.
Consider two sensors in air 15 cm apart, which is half a
wavelength for a 1000 Hz tone.

If we sum the signals of the mics with the same polarity,
we see that the system enhances a signal that impignes
from broadside (the direction perpendicular to the axis
between sensors) and reject signals impigning inline with
the axis bewteen sensors (can you see why?).

Similarly, if the signals are summed with reverse polarity,
the inline signal is permitted and the broadside signal
is rejected. Make a sketch on paper to see why.

Unfortunately, there are many complicating factors.
First, the beam width of the opening lobe is proportional
to the size of the array. Second, one has to make similar
considerations for lots of narrow frequency bands in order
to deal with broadband signals like speech. Third, multiple
reflections mess things seriously up if such a system
is used indoor.

> Compared to many of the current questions being asked this is very
> trivial.

The question may be trivial, but the answer is not...

> Many thanks in advance,

Y're welcome

> apologies if this seems like some daft questions,

By no means.

Rune

>
>flamingo11 skrev:
>> Hi, I'm not sure whether to be posting a new topic or piggybacking on
the
>> backs of many others, I've got so many questions  If people can point
me
>> at some literature that can point me in the right direction or have
some
>> comments about my proposed plan I'd be very grateful.
>
>> For my final year project I want to take a crude look at signal source
>> acquisition.
>
>Hmm... exactly what do you mean by "source acquisition"?
>Most people are familiar with "signal acquisition"...
>
>> My plan was to take some recordings using two cardiod
>> microphones and be able to 'listen' in certain directions, such as
towards
>> one of two people speaking.  I understand I can apply weighted delays
to
>> the incoming sounds to attenuate the surrounding signals from the one
I'm
>> trying to listen to.
>
>In principle, this is correct, yes.
>
>> In my case is it as simple as having two circular
>> buffers holding the incoming audio data and just applying a delay to
one
>> side or the other, then once a certain level of confidence has been
found
>> between the similarities of the two signals I can assume the beam is
being
>> steered in the direction I require?
>
>No. You could have locked on to that noisy AC unit humming
>in the corner of the room.
>
>> Once a certain level of confidence has
>> been achieved is it as simple as summing the two signals together to
create
>> a single mono channel with the required audio in it?
>
>Well... yes and no. The operations required to achieve
>beamforming are as simple as summing weighted signals.
>Consider two sensors in air 15 cm apart, which is half a
>wavelength for a 1000 Hz tone.
>
>If we sum the signals of the mics with the same polarity,
>we see that the system enhances a signal that impignes
>from broadside (the direction perpendicular to the axis
>between sensors) and reject signals impigning inline with
>the axis bewteen sensors (can you see why?).
>
>Similarly, if the signals are summed with reverse polarity,
>the inline signal is permitted and the broadside signal
>is rejected. Make a sketch on paper to see why.
>
>Unfortunately, there are many complicating factors.
>First, the beam width of the opening lobe is proportional
>to the size of the array. Second, one has to make similar
>considerations for lots of narrow frequency bands in order
>to deal with broadband signals like speech. Third, multiple
>reflections mess things seriously up if such a system
>is used indoor.
>
>> Compared to many of the current questions being asked this is very
>> trivial.
>
>The question may be trivial, but the answer is not...
>
>> Many thanks in advance,
>
>Y're welcome
>
>> apologies if this seems like some daft questions,
>
>By no means.
>
>Rune
>
>


Thanks for the reply Rune, looks like I've got a lot more research to get
involved with before doing much more with the project.  Do you know of any
books or material that would help me with my ideas?

Andrew

flamingo11 skrev:

> Thanks for the reply Rune, looks like I've got a lot more research to get
> involved with before doing much more with the project.  Do you know of any
> books or material that would help me with my ideas?
>
> Andrew

For beamforming alone, try Haykin's "Adaptive Filter Theory" (?).
For the application, you might want to check out various publications
in room acoustics. Room acoustics is a bit outside my main field
of interest so I don't know the details, but fairly recently there was
published a book on microphone arrays.

You might want to post on alt.sci.physics.acoustics.

Rune

>
>flamingo11 skrev:
>> Hi, I'm not sure whether to be posting a new topic or piggybacking on
the
>> backs of many others, I've got so many questions  If people can point
me
>> at some literature that can point me in the right direction or have
some
>> comments about my proposed plan I'd be very grateful.
>
>> For my final year project I want to take a crude look at signal source
>> acquisition.
>
>Hmm... exactly what do you mean by "source acquisition"?
>Most people are familiar with "signal acquisition"...
>
>> My plan was to take some recordings using two cardiod
>> microphones and be able to 'listen' in certain directions, such as
towards
>> one of two people speaking.  I understand I can apply weighted delays
to
>> the incoming sounds to attenuate the surrounding signals from the one
I'm
>> trying to listen to.
>
>In principle, this is correct, yes.
>
>> In my case is it as simple as having two circular
>> buffers holding the incoming audio data and just applying a delay to
one
>> side or the other, then once a certain level of confidence has been
found
>> between the similarities of the two signals I can assume the beam is
being
>> steered in the direction I require?
>
>No. You could have locked on to that noisy AC unit humming
>in the corner of the room.
>
>> Once a certain level of confidence has
>> been achieved is it as simple as summing the two signals together to
create
>> a single mono channel with the required audio in it?
>
>Well... yes and no. The operations required to achieve
>beamforming are as simple as summing weighted signals.
>Consider two sensors in air 15 cm apart, which is half a
>wavelength for a 1000 Hz tone.
>
>If we sum the signals of the mics with the same polarity,
>we see that the system enhances a signal that impignes
>from broadside (the direction perpendicular to the axis
>between sensors) and reject signals impigning inline with
>the axis bewteen sensors (can you see why?).
>
>Similarly, if the signals are summed with reverse polarity,
>the inline signal is permitted and the broadside signal
>is rejected. Make a sketch on paper to see why.
>
>Unfortunately, there are many complicating factors.
>First, the beam width of the opening lobe is proportional
>to the size of the array. Second, one has to make similar
>considerations for lots of narrow frequency bands in order
>to deal with broadband signals like speech. Third, multiple
>reflections mess things seriously up if such a system
>is used indoor.
>
>> Compared to many of the current questions being asked this is very
>> trivial.
>
>The question may be trivial, but the answer is not...
>
>> Many thanks in advance,
>
>Y're welcome
>
>> apologies if this seems like some daft questions,
>
>By no means.
>
>Rune
>
>


Thanks for the reply Rune, looks like I've got a lot more research to get
involved with before doing much more with the project.  Do you know of any
books or material that would help me with my ideas?

Andrew