>> No worries. So the next group of
three talks, then we'll have another
break, starting with Bjorn Roche, and
the title of this talk is "How I used
knowledge of code and music to fight
fires!". I can't read my own writing. Then
it's Chloe Weil with "Tasty Stacks!", and I
don't know if you've read this abstract,
but the title is... There's a little
more to it. And then it's Julia Evans
with "Spying on programs using strace!!!". So
without further ado, Bjorn Roche.

 >> Hi! Okay. So that is my talk title,
and this is my official talk title. With
all the requisite... So I'd like to
start by talking a little bit about
myself. This morning, we had a talk
about obsession. And I am obsessed... I
am a contract developer who's obsessed
with music and audio. And DSP, digital
signal processing. So I've developed
software for, like, music startups,
recording studios, that kind of thing.
That's what I do. I'm going to skip this
other stuff, except I will mention my
current project is an app that allows
users to create music videos. It's
really cool. It's called Shimmeo. So I
was surprised when this company called
edispatches contacted me about their
thing for firefighters. So this talk is
about what I did for them.

So the way... Is anybody here a first
responder or has been a first responder,
like a volunteer firefighter, EMS,
anything like that? No. Okay. Then you
won't know if I got anything wrong, I
guess. So... The way first responders
respond to emergencies in a lot of
areas... I don't think this happens in
big cities, but I don't really know...
Is the fire station receives a call,
they need to send out first responders,
so they have a radio station inside the
fire station that sends out a message,
people are carrying pagers, they're
called pagers. They're not really
pagers. They're really radios. And they
hear a little message that says...

(static)

It's about as clear as the New York City
subway used to be. And somehow they
understand what it says, and they run
off to the fire or whatever it is. The
messages themselves are audio messages.
They start with tone sets. That sound
like...

(singing)

Doo-deet. And those tone sets encode for
each pager. So the tones I just made
might be encoded for the green pagers,
and there might be other tone sets that
are encoded for other pagers. So when
the tone set goes, then a particular
pager says -- oh, that's me, and turns
on. So the green pager turns on when it
hears the green tone set, and it'll
actually -- the person with that pager
will actually hear that blue tone set,
and whatever else happens, and then
they'll hear the message. So it's
actually a little speaker. It's not text
or anything fancy. They carry a little
pager with a speaker on it, and they
have to actually hear that message. And
I have a sample here, which... You'll
probably hear.

(tone sets)

That's the first tone set. So what you
just heard part of was two tone sets,
somebody talking, and then some noise,
which is something you always get. So
the system works. And they've been doing
it for years this way, all over the
country, all over -- I don't know where
else, but at least North America. But
the problem is people need to carry
these pagers. So you've got your phone,
you've got your pager. The pager uses a
lot of battery, so you need to charge it
regularly, remember to charge it more
than your phone. It has a limited range.

The messages are just played once, so
you hear the message, and if you're a
volunteer firefighter, and also a
construction worker, and you miss it, or
you didn't understand one, two, three
evergreen terrace... Is that what they
said? You're out of luck. You don't know
where to go. And many first responders
are also construction workers and stuff
like that. So... And another problem is
they need to buy all these extra pagers,
and each pager needs to be programmed
for the specific message, so the
individual who has it gets the right
message. So it would be pretty awesome
if we could just say... Hey, fire
station, send out a message to the
internet, and the internet will
magically deploy to all the right
places.

But in the real world, we don't live
that way. Everybody is really invested
in this radio system deployment. So the
edispatches folks figured out that it
would be better than reinventing the
wheel -- to attach a new wheel, so they
developed this system that I'm going to
show you and that I worked on, to
connect -- basically they added a second
radio receiver that acts like a pager,
and something they call a detector, but
is actually this little computer here.
We used to call these embedded
computers. What do you call them now?

I don't know. Is there a name for this?
Embedded computers. Okay. Still call
them that? All right. So it's just a
Linux box with some... With an audio
input on one side, and Ethernet on the
other, and it has a bunch of other
things that don't get used, and here's
the radio receiver. It's just a standard
short wave radio receiver. And so that's
what the radio receiver and detector
are, and then the wire here that you're
all familiar with, the Ethernet wire,
goes to the internet, and they have a
system that connects it all to
everybody's phones.

So the key, really, that I worked on, is
the detector part, which I'll talk a
little bit about, and yes, I still have
five minutes! Wow. And so the way that
works... These are the main parts. The
main parts of the software. And I don't
have to talk about the first step,
because there was another talk earlier
today all about reading audio, using
PortAudio. Which is really awesome. So
that's how the analog input gets read.
It's just audio. It's just sound. That's
what's coming out of the radio. The next
step is... Automatic gain control.
Basically this detects any distortion
that might come from a level mismatch
between the two devices.

And adjusts the input level as
necessary. This was one of the easiest
things I did, but it probably made the
biggest difference. Because before they
had a special cable with an attenuator
in it, and it, like, adjusted... And oh,
that seems to work. Well... It seems to
work better if we turn it two times
around, rather than three. And it was
very haphazard. And then there are some
filters. So if you've ever played... Who
here has played guitar? Yes. Okay. So no
first responders, but lots of
guitarists. So if you've played guitar,
you've played guitar through an
amplifier. Like... We've got one over
there. Is that a Fender? Yes. And the
amplifier distorts, so the sound coming
out of the guitar sounds nice and pure.
Doot! But when it goes to the amp, it
sounds like... Durrr! And we like that
sound, for whatever reason. Our ears
love that sound.

And what that sound is caused by is...
Extra frequencies being added, that are
higher frequency than the original. So
what this filter does is reduce some of
that high frequency distortion that
comes through the radio, just like it
does through the guitar amp. It's a
natural process. And the next step is...
We'll really determine what frequencies
are there. How much 200 hertz, how much
210 hertz, how much 220 hertz, and so
on. And we pass that information on to
something I call the analyzer, which
keeps a history of all that information,
and figures out if an actual tone has
been played. So there's some statistics
going on there. But it's not too
complicated.

And then if an actual tone has been
played, it starts recording, and sends
the message off to the lovely internets.
All right. So we're going to do the
demo real quick. And this is the part of
the system that I'm actually going to be
demonstrating. So let's see. So this
here, my computer is my fire station.
It's going to play the tone. And the
message. This is my transmitter. My
fancy radio station. This is the
receiver. And then this is the detector.
And then you'll see the output of the
detector up here. And if you blink,
you'll miss it, so don't blink.

What you'll see is the gain... Well,
I'll show you the results, and I'll do
it again. Because it's so quick. All
right.

(tones)

 >> The first tone.

 >> This is HSH. This is a test.

 >> Hopefully it got both of them. Yeah.
So you can see the first line there, the
automatic gain control says -- whoa,
that's too loud. It turns it down a
little bit. And then hit one, it detects
those first set of tones, which are 1901
hertz, followed by 350 hertz, and then
it tries to notify the server, but we're
not connected to the internet. And then
the second one -- it detects the second
set of tones, and that time it's 350,
followed by 1985. And I'll do it one
more time, so that now you know what's
going on, you can see it.

(tones)

(this is HSH. This is a test.)

 >> And those tones sound so beautiful.
So that's how the system works. If you
want to know more, I know I'm out of
time... You want to know more, here's
some stuff about -- the green parts of
the diagram there are Open Source. They
came from a guitar tuner I made. And
there's more info here.

(applause)