WEBVTT

00:00:00.000 --> 00:00:05.100 align:middle line:90%


00:00:05.100 --> 00:00:07.140 align:middle line:84%
Now we have been looking
at different types

00:00:07.140 --> 00:00:10.350 align:middle line:84%
of motion capture, where we
can put markers on the body

00:00:10.350 --> 00:00:14.850 align:middle line:84%
and capture the motion of
people with a very high level

00:00:14.850 --> 00:00:16.470 align:middle line:90%
of accuracy and precision.

00:00:16.470 --> 00:00:18.420 align:middle line:84%
And that's often
the preferable way

00:00:18.420 --> 00:00:23.280 align:middle line:84%
of doing it when we're doing
the type of studies we do.

00:00:23.280 --> 00:00:25.260 align:middle line:84%
But I would also like
to mention that there

00:00:25.260 --> 00:00:29.190 align:middle line:84%
are many also good reasons for
doing regular video recordings

00:00:29.190 --> 00:00:32.189 align:middle line:90%
and do analysis based on video.

00:00:32.189 --> 00:00:36.090 align:middle line:84%
And first of all, one of the
nice things about video cameras

00:00:36.090 --> 00:00:38.280 align:middle line:90%
is that they are everywhere.

00:00:38.280 --> 00:00:41.830 align:middle line:84%
You have them in
your mobile phone.

00:00:41.830 --> 00:00:44.460 align:middle line:84%
So from this you can actually
get a quite nice video

00:00:44.460 --> 00:00:45.630 align:middle line:90%
recording.

00:00:45.630 --> 00:00:47.970 align:middle line:84%
At least if you
put it on a tripod,

00:00:47.970 --> 00:00:51.150 align:middle line:84%
so the camera doesn't
move very much.

00:00:51.150 --> 00:00:53.530 align:middle line:84%
You can also get all sorts
of other types of cameras

00:00:53.530 --> 00:00:54.030 align:middle line:90%
these days.

00:00:54.030 --> 00:00:56.370 align:middle line:10%
For example, such portable
action cameras that

00:00:56.370 --> 00:01:00.000 align:middle line:10%
are quite nice to put up in
different constellations.

00:01:00.000 --> 00:01:02.640 align:middle line:84%
I often like to put them in the
ceiling above a stage, where

00:01:02.640 --> 00:01:04.920 align:middle line:84%
it's possible to look at
how performers, for example,

00:01:04.920 --> 00:01:05.940 align:middle line:90%
move on stage.

00:01:05.940 --> 00:01:07.950 align:middle line:84%
Or you can put them in
all the small spots where

00:01:07.950 --> 00:01:10.260 align:middle line:84%
it's difficult to get
through to be able to look

00:01:10.260 --> 00:01:11.220 align:middle line:90%
at what's going on.

00:01:11.220 --> 00:01:14.580 align:middle line:84%
For example, like looking
at the feet of a pianist.

00:01:14.580 --> 00:01:17.040 align:middle line:84%
Of course, you can also
scale up and have a more,

00:01:17.040 --> 00:01:18.660 align:middle line:84%
kind of a little
bit larger camera

00:01:18.660 --> 00:01:21.820 align:middle line:84%
that you can put at
the front of a stage.

00:01:21.820 --> 00:01:24.850 align:middle line:84%
Or even go up to more kind of
professional type of cameras,

00:01:24.850 --> 00:01:27.750 align:middle line:84%
like this one, if you really
want to get good results.

00:01:27.750 --> 00:01:32.280 align:middle line:84%
But for many purposes, you
can get very good results

00:01:32.280 --> 00:01:35.590 align:middle line:84%
out of just using a
small little camera.

00:01:35.590 --> 00:01:39.368 align:middle line:84%
So when it comes to
recording video for analysis,

00:01:39.368 --> 00:01:41.160 align:middle line:84%
that's quite different
than recording video

00:01:41.160 --> 00:01:43.860 align:middle line:84%
for aesthetic
reasons or for doing

00:01:43.860 --> 00:01:45.390 align:middle line:90%
other types of production.

00:01:45.390 --> 00:01:48.810 align:middle line:84%
Because usually then people
would move the cameras a lot,

00:01:48.810 --> 00:01:51.360 align:middle line:90%
zoom in, pan, tilt, et cetera.

00:01:51.360 --> 00:01:55.170 align:middle line:84%
But when you're doing
analysis on video material,

00:01:55.170 --> 00:01:56.880 align:middle line:84%
that type of movement
of the camera

00:01:56.880 --> 00:01:58.860 align:middle line:84%
is not very good,
because that will show up

00:01:58.860 --> 00:02:01.330 align:middle line:90%
in the video analysis.

00:02:01.330 --> 00:02:05.610 align:middle line:10%
So when we shoot video
for analysis purposes,

00:02:05.610 --> 00:02:10.110 align:middle line:10%
we always try to have the
camera stand still on the tripod

00:02:10.110 --> 00:02:13.470 align:middle line:10%
or on a table or some
other kind of stand,

00:02:13.470 --> 00:02:15.090 align:middle line:10%
so that it's possible
to only capture

00:02:15.090 --> 00:02:18.480 align:middle line:10%
the motion within the frame and
not the movement of the camera

00:02:18.480 --> 00:02:19.290 align:middle line:10%
itself.

00:02:19.290 --> 00:02:22.050 align:middle line:84%
We also try to have,
then, good light

00:02:22.050 --> 00:02:24.480 align:middle line:84%
and also try to remove
the background as much

00:02:24.480 --> 00:02:26.430 align:middle line:84%
as possible so that
it's easy to look

00:02:26.430 --> 00:02:27.870 align:middle line:90%
at what is in the foreground.

00:02:27.870 --> 00:02:31.800 align:middle line:84%
For example, a musician
or a dancer on stage.

00:02:31.800 --> 00:02:35.340 align:middle line:84%
So then if you have made a
recording that you think is

00:02:35.340 --> 00:02:38.310 align:middle line:84%
nice and that you want
to use for analysis,

00:02:38.310 --> 00:02:41.460 align:middle line:84%
then there are different
approaches to how to analyse

00:02:41.460 --> 00:02:42.740 align:middle line:90%
this.

00:02:42.740 --> 00:02:45.040 align:middle line:84%
And one of them is
that of doing a more

00:02:45.040 --> 00:02:49.410 align:middle line:84%
of a qualitative analysis,
which is based on observation.

00:02:49.410 --> 00:02:52.620 align:middle line:84%
And in that case, you would
take a look at the video file,

00:02:52.620 --> 00:02:55.350 align:middle line:84%
and just by your eyes,
observe what is going on.

00:02:55.350 --> 00:02:58.230 align:middle line:84%
For example, also notate down,
or use a computer programme

00:02:58.230 --> 00:03:03.850 align:middle line:84%
to help in making a notation
of the video in question.

00:03:03.850 --> 00:03:06.960 align:middle line:84%
The next step could be to do
more of a quantitative type

00:03:06.960 --> 00:03:10.740 align:middle line:84%
of analysis, where you run
the video through a computer

00:03:10.740 --> 00:03:13.800 align:middle line:84%
programme that will calculate
different types of features.

00:03:13.800 --> 00:03:15.780 align:middle line:84%
And it can calculate,
for example, the quantity

00:03:15.780 --> 00:03:21.120 align:middle line:84%
of motion, that is, how much a
person is moving in the frame.

00:03:21.120 --> 00:03:23.760 align:middle line:84%
Look at where in the
frame a person is moving.

00:03:23.760 --> 00:03:26.490 align:middle line:84%
For example, kind of up
and down or sideways.

00:03:26.490 --> 00:03:29.260 align:middle line:90%
And other the types of features.

00:03:29.260 --> 00:03:31.530 align:middle line:84%
So that's kind of the
qualitative and quantitative

00:03:31.530 --> 00:03:33.210 align:middle line:84%
type of approaches
to video analysis,

00:03:33.210 --> 00:03:34.843 align:middle line:84%
but there's also
something in between

00:03:34.843 --> 00:03:36.510 align:middle line:84%
that we have been
working on quite a lot

00:03:36.510 --> 00:03:37.980 align:middle line:90%
here at the University of Oslo.

00:03:37.980 --> 00:03:41.050 align:middle line:10%
And that is what we call
video visualisation.

00:03:41.050 --> 00:03:44.220 align:middle line:10%
And this is a way of doing
a kind of quantitative type

00:03:44.220 --> 00:03:49.200 align:middle line:10%
of analysis on the video, but
made for qualitative purposes.

00:03:49.200 --> 00:03:52.920 align:middle line:84%
And one such example, here,
is what I have behind me here.

00:03:52.920 --> 00:03:55.830 align:middle line:84%
And this is what I
call a motiongram.

00:03:55.830 --> 00:03:58.560 align:middle line:10%
And this is a very
compact representation

00:03:58.560 --> 00:04:01.110 align:middle line:10%
of a movement sequence.

00:04:01.110 --> 00:04:04.440 align:middle line:84%
In this case, it's the
movement of a dancer.

00:04:04.440 --> 00:04:07.850 align:middle line:84%
And this is the
original video file.

00:04:07.850 --> 00:04:13.160 align:middle line:84%
So we see here that a dancer,
she's moving her arms--

00:04:13.160 --> 00:04:16.730 align:middle line:84%
And this is from a study we did
looking at spontaneous dance

00:04:16.730 --> 00:04:18.320 align:middle line:90%
movements to music.

00:04:18.320 --> 00:04:20.600 align:middle line:84%
And, just by looking
at this, we get

00:04:20.600 --> 00:04:26.060 align:middle line:84%
a sense of how the movement is
unfolding in time and space.

00:04:26.060 --> 00:04:28.760 align:middle line:84%
But it's difficult
to really grasp

00:04:28.760 --> 00:04:30.390 align:middle line:90%
how it looks like over time.

00:04:30.390 --> 00:04:33.510 align:middle line:84%
So, for example, if you want to
put this into a research paper,

00:04:33.510 --> 00:04:36.930 align:middle line:84%
we need to capture this
in one way or another.

00:04:36.930 --> 00:04:39.680 align:middle line:84%
And that's when these
motiongrams are useful.

00:04:39.680 --> 00:04:42.920 align:middle line:84%
Because the motiongram is a
representation of the movement

00:04:42.920 --> 00:04:43.850 align:middle line:90%
we just saw.

00:04:43.850 --> 00:04:47.250 align:middle line:84%
We can see the hands, her
hands, moving up and down here.

00:04:47.250 --> 00:04:49.670 align:middle line:84%
Here she's standing
more or less still.

00:04:49.670 --> 00:04:51.990 align:middle line:84%
And then here she's
moving up again.

00:04:51.990 --> 00:04:55.280 align:middle line:84%
So it's kind of a
way of representing

00:04:55.280 --> 00:04:58.880 align:middle line:90%
the motion over time.

00:04:58.880 --> 00:05:01.130 align:middle line:10%
This is just one of
many different types

00:05:01.130 --> 00:05:04.070 align:middle line:10%
of video visualisations.

00:05:04.070 --> 00:05:07.400 align:middle line:84%
One of the reasons it's possible
to do any type of analysis

00:05:07.400 --> 00:05:11.240 align:middle line:84%
of a video file to start with
is that a video file is actually

00:05:11.240 --> 00:05:13.020 align:middle line:90%
just a series of numbers.

00:05:13.020 --> 00:05:15.530 align:middle line:84%
So it's a kind of
a matrix, where

00:05:15.530 --> 00:05:19.130 align:middle line:84%
you have the grid, where you
have the pixels in the image.

00:05:19.130 --> 00:05:24.140 align:middle line:84%
And then you have four layers
in each of the frames that will

00:05:24.140 --> 00:05:26.720 align:middle line:84%
correspond to the different
colours in the image.

00:05:26.720 --> 00:05:29.490 align:middle line:84%
So based on these numbers,
it's possible to calculate,

00:05:29.490 --> 00:05:31.820 align:middle line:84%
for example, the
average of what is going

00:05:31.820 --> 00:05:34.460 align:middle line:90%
on in one of these images.

00:05:34.460 --> 00:05:36.920 align:middle line:84%
And one technique
that we often use

00:05:36.920 --> 00:05:41.300 align:middle line:84%
is to create what we
call a motion image.

00:05:41.300 --> 00:05:44.510 align:middle line:84%
And here it is that we will
start from a normal video

00:05:44.510 --> 00:05:48.080 align:middle line:84%
recording, then we can crop it
a little bit so that you get

00:05:48.080 --> 00:05:49.672 align:middle line:90%
kind of closer to the image.

00:05:49.672 --> 00:05:52.130 align:middle line:84%
We can change, for example,
the brightness and the contrast

00:05:52.130 --> 00:05:54.950 align:middle line:84%
so it's easier to separate the
foreground from the background.

00:05:54.950 --> 00:05:58.180 align:middle line:84%
And then we can calculate what
we call the frame difference.

00:05:58.180 --> 00:06:00.800 align:middle line:10%
So we take one frame
and then the next frame

00:06:00.800 --> 00:06:03.260 align:middle line:10%
and we subtract them
mathematically from each other.

00:06:03.260 --> 00:06:05.090 align:middle line:10%
And what you end
up with is what we

00:06:05.090 --> 00:06:08.900 align:middle line:10%
call a motion image that
is showing what changed

00:06:08.900 --> 00:06:10.760 align:middle line:10%
between these two frames.

00:06:10.760 --> 00:06:15.980 align:middle line:84%
And then that is something
you can look at as a video,

00:06:15.980 --> 00:06:18.020 align:middle line:84%
so you can see only the
parts in the image that

00:06:18.020 --> 00:06:19.240 align:middle line:90%
changed over time.

00:06:19.240 --> 00:06:23.270 align:middle line:84%
And this is a very common way
to start to do an analysis.

00:06:23.270 --> 00:06:24.650 align:middle line:84%
And then from that
one, again, we

00:06:24.650 --> 00:06:27.110 align:middle line:84%
can create what we
call this motiongram.

00:06:27.110 --> 00:06:31.310 align:middle line:84%
That's this representation
over time of a motion sequence

00:06:31.310 --> 00:06:33.290 align:middle line:84%
that's based kind of
squeezing together

00:06:33.290 --> 00:06:36.980 align:middle line:84%
each of the motion images
and plotting them over time.

00:06:36.980 --> 00:06:40.640 align:middle line:10%
It's also possible to look at
what we call a motion history

00:06:40.640 --> 00:06:44.630 align:middle line:10%
video, where you can
see kind of traces

00:06:44.630 --> 00:06:48.230 align:middle line:10%
of the motion of a person
over time as we see here.

00:06:48.230 --> 00:06:51.050 align:middle line:10%
With different types
of philtres on top.

00:06:51.050 --> 00:06:54.320 align:middle line:10%
So this we can kind of tailor
to the particular motion

00:06:54.320 --> 00:06:55.400 align:middle line:10%
in question.

00:06:55.400 --> 00:06:57.850 align:middle line:10%
And together then,
a motiongram is

00:06:57.850 --> 00:06:59.600 align:middle line:10%
kind of a representation
that can give you

00:06:59.600 --> 00:07:03.830 align:middle line:10%
a sense of how the body is
moving in time and space.

00:07:03.830 --> 00:07:05.690 align:middle line:10%
While in this case,
here, we can look

00:07:05.690 --> 00:07:12.980 align:middle line:10%
at a continuous video, where we
see how the motion is changing

00:07:12.980 --> 00:07:14.600 align:middle line:10%
spatially.

00:07:14.600 --> 00:07:16.513 align:middle line:10%
And a course, based
on this again,

00:07:16.513 --> 00:07:18.680 align:middle line:10%
it's possible to calculate
various types of features

00:07:18.680 --> 00:07:21.490 align:middle line:10%
that can be used in
quantitative measurements.

00:07:21.490 --> 00:07:24.800 align:middle line:10%
So we often like to work
within both quantitative and

00:07:24.800 --> 00:07:27.770 align:middle line:10%
qualitative video
analysis approaches

00:07:27.770 --> 00:07:30.140 align:middle line:10%
and use these types of
visualisation techniques

00:07:30.140 --> 00:07:34.940 align:middle line:10%
as a guide to help us with
finding what is important.

00:07:34.940 --> 00:07:39.200 align:middle line:84%
So it's possible to do this type
of video analysis on its own,

00:07:39.200 --> 00:07:42.110 align:middle line:84%
but we often combine
it with motion capture

00:07:42.110 --> 00:07:44.780 align:middle line:84%
so that we can get the
best of both worlds.

00:07:44.780 --> 00:07:47.150 align:middle line:84%
That is, a very high level
of accuracy and precision

00:07:47.150 --> 00:07:49.160 align:middle line:84%
from the motion capture
system, and then

00:07:49.160 --> 00:07:55.240 align:middle line:84%
a more holistic and global
view from the video recordings.

00:07:55.240 --> 00:08:01.000 align:middle line:90%