WEBVTT

00:00:00.000 --> 00:00:05.220 align:middle line:90%


00:00:05.220 --> 00:00:12.570 align:middle line:84%
So in-- some systems have
pre-defined skeletons that you

00:00:12.570 --> 00:00:17.190 align:middle line:84%
can use, but in this programme
that we're using now,

00:00:17.190 --> 00:00:23.310 align:middle line:84%
you can actually create bones
yourself, between the markers,

00:00:23.310 --> 00:00:26.180 align:middle line:84%
which makes it easier to see
what you're actually looking

00:00:26.180 --> 00:00:26.680 align:middle line:90%
at.

00:00:26.680 --> 00:00:28.740 align:middle line:90%
And here is a recording.

00:00:28.740 --> 00:00:34.005 align:middle line:84%
and where the markers
are attached by bones.

00:00:34.005 --> 00:00:36.820 align:middle line:90%


00:00:36.820 --> 00:00:38.400 align:middle line:90%
Yes.

00:00:38.400 --> 00:00:42.390 align:middle line:84%
Probably it's also
useful to explain,

00:00:42.390 --> 00:00:47.340 align:middle line:84%
if you choose to capture
people using a skeleton,

00:00:47.340 --> 00:00:52.620 align:middle line:84%
this is kind of a
model that the software

00:00:52.620 --> 00:00:56.670 align:middle line:84%
uses that helps it keep track
of the different markers

00:00:56.670 --> 00:00:58.440 align:middle line:90%
and where they are.

00:00:58.440 --> 00:01:02.300 align:middle line:84%
So there are different skeletons
that you can choose from.

00:01:02.300 --> 00:01:03.800 align:middle line:84%
And then you have
to put the markers

00:01:03.800 --> 00:01:07.640 align:middle line:84%
in that specific
location, according

00:01:07.640 --> 00:01:09.800 align:middle line:84%
to the way the
skeleton is defined.

00:01:09.800 --> 00:01:12.890 align:middle line:90%
And this should help.

00:01:12.890 --> 00:01:15.783 align:middle line:84%
It should help with tracking
if you put the markers

00:01:15.783 --> 00:01:16.700 align:middle line:90%
in the right location.

00:01:16.700 --> 00:01:19.790 align:middle line:84%
Because the system knows,
if the elbow disappears,

00:01:19.790 --> 00:01:23.540 align:middle line:84%
it knows to look again
in this location, when

00:01:23.540 --> 00:01:25.220 align:middle line:90%
another marker reappears.

00:01:25.220 --> 00:01:29.180 align:middle line:84%
And hopefully then it joins
those disjointed trajectories

00:01:29.180 --> 00:01:33.560 align:middle line:84%
together and saves you some
time with the labelling.

00:01:33.560 --> 00:01:35.150 align:middle line:84%
And the other thing
that you can do

00:01:35.150 --> 00:01:41.810 align:middle line:84%
is define an object
as a rigid object.

00:01:41.810 --> 00:01:46.250 align:middle line:84%
So for me, I might do this
with a musical instrument.

00:01:46.250 --> 00:01:48.590 align:middle line:84%
And you require a few
markers to do this.

00:01:48.590 --> 00:01:51.680 align:middle line:84%
And this would be
in the case where

00:01:51.680 --> 00:01:54.080 align:middle line:84%
the markers do not
change their position

00:01:54.080 --> 00:01:55.590 align:middle line:90%
relative to each other.

00:01:55.590 --> 00:01:57.980 align:middle line:84%
So it has to be
something that's solid.

00:01:57.980 --> 00:01:59.810 align:middle line:90%
So if you imagine a violin.

00:01:59.810 --> 00:02:01.760 align:middle line:84%
You can put a few
markers on a violin

00:02:01.760 --> 00:02:04.250 align:middle line:84%
and it's not that you're going
to break the violin in half,

00:02:04.250 --> 00:02:07.190 align:middle line:84%
so the markers position
relative to each other

00:02:07.190 --> 00:02:08.560 align:middle line:90%
will stay the same.

00:02:08.560 --> 00:02:11.750 align:middle line:84%
This is different from a
person, where the person moves

00:02:11.750 --> 00:02:13.050 align:middle line:90%
in all kinds of ways.

00:02:13.050 --> 00:02:16.460 align:middle line:84%
So the markers of the
elbow are sometimes

00:02:16.460 --> 00:02:20.180 align:middle line:84%
going to be closer to the head,
and sometimes further away.

00:02:20.180 --> 00:02:27.320 align:middle line:84%
So when you're happy with your
labelling and gap-filling,

00:02:27.320 --> 00:02:29.600 align:middle line:84%
then you need to
export your data.

00:02:29.600 --> 00:02:33.590 align:middle line:10%
And you do that
into a text file.

00:02:33.590 --> 00:02:37.310 align:middle line:10%
And the format of that file, it
kind of depends on the software

00:02:37.310 --> 00:02:41.100 align:middle line:10%
that you will use later
for analysing your data.

00:02:41.100 --> 00:02:43.700 align:middle line:90%


00:02:43.700 --> 00:02:46.600 align:middle line:84%
And so you have many choices
when you export the data,

00:02:46.600 --> 00:02:48.850 align:middle line:84%
depending on what
you want to take

00:02:48.850 --> 00:02:51.080 align:middle line:90%
and what you want to do with it.

00:02:51.080 --> 00:02:56.430 align:middle line:84%
So you might choose to just
export marker positions,

00:02:56.430 --> 00:03:00.240 align:middle line:84%
or you maybe you've defined
bones between the markers,

00:03:00.240 --> 00:03:04.830 align:middle line:84%
and so you want to also export
information about the bones.

00:03:04.830 --> 00:03:09.240 align:middle line:84%
So the systems can use
algorithms to estimate

00:03:09.240 --> 00:03:11.880 align:middle line:84%
the centre of a
bone or something.

00:03:11.880 --> 00:03:16.140 align:middle line:84%
With rigid objects, you can
also export information about

00:03:16.140 --> 00:03:19.710 align:middle line:84%
the centre of the object in
the way it's rotated in space.

00:03:19.710 --> 00:03:22.600 align:middle line:84%
So once you've exported
your data as a text file

00:03:22.600 --> 00:03:25.590 align:middle line:84%
and you've imported it into
whatever software you like

00:03:25.590 --> 00:03:29.790 align:middle line:84%
to analyse things in, usually
the next step would be to have

00:03:29.790 --> 00:03:33.670 align:middle line:84%
a look at your data and see if
you need to do any smoothing.

00:03:33.670 --> 00:03:35.280 align:middle line:84%
So we've already
talked a little bit

00:03:35.280 --> 00:03:38.550 align:middle line:84%
about the type of noise that
can occur in motion capture

00:03:38.550 --> 00:03:40.260 align:middle line:90%
recordings.

00:03:40.260 --> 00:03:47.150 align:middle line:84%
This plot shows in black
the original recording

00:03:47.150 --> 00:03:50.270 align:middle line:84%
of a particular
marker, and you can

00:03:50.270 --> 00:03:53.430 align:middle line:84%
see that there's a small
amounts of jitter happening.

00:03:53.430 --> 00:03:58.100 align:middle line:84%
So this is really zoomed in on
a very small part of the marker

00:03:58.100 --> 00:03:59.150 align:middle line:90%
trajectory.

00:03:59.150 --> 00:04:03.970 align:middle line:90%
This is 200 times--

00:04:03.970 --> 00:04:08.140 align:middle line:84%
200 times 8, so it's
less than two seconds

00:04:08.140 --> 00:04:09.850 align:middle line:90%
that we're seeing here.

00:04:09.850 --> 00:04:13.780 align:middle line:84%
And it might be useful for
you to get rid of this jitter

00:04:13.780 --> 00:04:15.350 align:middle line:90%
that you see in the black line.

00:04:15.350 --> 00:04:16.790 align:middle line:90%
So this is what I've done.

00:04:16.790 --> 00:04:20.200 align:middle line:84%
You can use some sort of
mathematical smoothing

00:04:20.200 --> 00:04:20.800 align:middle line:90%
algorithm.

00:04:20.800 --> 00:04:24.520 align:middle line:10%
There are various choices
to get rid of this jitter.

00:04:24.520 --> 00:04:27.340 align:middle line:84%
So the red line here
is a smoothed version

00:04:27.340 --> 00:04:29.560 align:middle line:90%
of that marker trajectory.

00:04:29.560 --> 00:04:33.790 align:middle line:84%
And it follows the same
contours and maintains

00:04:33.790 --> 00:04:35.860 align:middle line:84%
the important
information, but it just

00:04:35.860 --> 00:04:38.500 align:middle line:84%
gets rid of all the
small amounts of jitter.

00:04:38.500 --> 00:04:41.380 align:middle line:84%
And I would say this is
especially important if you're

00:04:41.380 --> 00:04:46.150 align:middle line:84%
planning to look at
derivatives of position,

00:04:46.150 --> 00:04:50.690 align:middle line:84%
like velocity or acceleration,
because whatever jitter there

00:04:50.690 --> 00:04:54.340 align:middle line:84%
is there, it's going to
multiply itself every time you

00:04:54.340 --> 00:04:57.250 align:middle line:90%
take a derivative.

00:04:57.250 --> 00:05:01.330 align:middle line:84%
So it's important, if you don't
want to look at that noise,

00:05:01.330 --> 00:05:02.820 align:middle line:84%
it's important to
get rid of that.

00:05:02.820 --> 00:05:05.820 align:middle line:90%


00:05:05.820 --> 00:05:07.670 align:middle line:90%
That good?

00:05:07.670 --> 00:05:16.000 align:middle line:90%