TEDxBeaconStreet
Michael Rubinstein: See invisible motion, hear silent sounds
Filmed:
Readability: 4.1
2,075,056 views
Meet the “motion microscope,” a video-processing tool that plays up tiny changes in motion and color impossible to see with the naked eye. Video researcher Michael Rubinstein plays us clip after jaw-dropping clip showing how this tech can track an individual’s pulse and heartbeat simply from a piece of footage. Watch him re-create a conversation by amplifying the movements from sound waves bouncing off a bag of chips. The wow-inspiring and sinister applications of this tech you have to see to believe.
Michael Rubinstein - Research scientist, Google
Computer scientist Michael Rubinstein and his team have developed a "motion microscope" that can show video footage of barely perceivable movements, like breaths and heartbeats. Full bio
Computer scientist Michael Rubinstein and his team have developed a "motion microscope" that can show video footage of barely perceivable movements, like breaths and heartbeats. Full bio
Double-click the English transcript below to play the video.
00:13
So over the past few centuries,
microscopes have revolutionized our world.
microscopes have revolutionized our world.
0
1114
7492
00:21
They revealed to us a tiny world
of objects, life and structures
of objects, life and structures
1
9036
5216
00:26
that are too small for us
to see with our naked eyes.
to see with our naked eyes.
2
14252
2906
00:29
They are a tremendous contribution
to science and technology.
to science and technology.
3
17158
3019
00:32
Today I'd like to introduce you
to a new type of microscope,
to a new type of microscope,
4
20177
3227
00:35
a microscope for changes.
5
23404
2578
00:37
It doesn't use optics
like a regular microscope
like a regular microscope
6
25982
2902
00:40
to make small objects bigger,
7
28884
1997
00:42
but instead it uses a video camera
and image processing
and image processing
8
30881
4376
00:47
to reveal to us the tiniest motions
and color changes in objects and people,
and color changes in objects and people,
9
35257
5256
00:52
changes that are impossible
for us to see with our naked eyes.
for us to see with our naked eyes.
10
40513
3842
00:56
And it lets us look at our world
in a completely new way.
in a completely new way.
11
44355
4120
01:00
So what do I mean by color changes?
12
48475
1910
01:02
Our skin, for example,
changes its color very slightly
changes its color very slightly
13
50385
2832
01:05
when the blood flows under it.
14
53217
1997
01:07
That change is incredibly subtle,
15
55214
2397
01:09
which is why, when you
look at other people,
look at other people,
16
57611
2063
01:11
when you look at the person
sitting next to you,
sitting next to you,
17
59674
2251
01:13
you don't see their skin
or their face changing color.
or their face changing color.
18
61925
3575
01:17
When we look at this video of Steve here,
it appears to us like a static picture,
it appears to us like a static picture,
19
65500
4360
01:21
but once we look at this video
through our new, special microscope,
through our new, special microscope,
20
69860
3860
01:25
suddenly we see
a completely different image.
a completely different image.
21
73720
2600
01:28
What you see here are small changes
in the color of Steve's skin,
in the color of Steve's skin,
22
76320
3930
01:32
magnified 100 times
so that they become visible.
so that they become visible.
23
80250
4436
01:36
We can actually see a human pulse.
24
84686
3267
01:39
We can see how fast
Steve's heart is beating,
Steve's heart is beating,
25
87953
3227
01:43
but we can also see the actual way
that the blood flows in his face.
that the blood flows in his face.
26
91180
5355
01:48
And we can do that not just
to visualize the pulse,
to visualize the pulse,
27
96544
2631
01:51
but also to actually
recover our heart rates,
recover our heart rates,
28
99175
3471
01:54
and measure our heart rates.
29
102646
1793
01:56
And we can do it with regular cameras
and without touching the patients.
and without touching the patients.
30
104439
4453
02:00
So here you see the pulse and heart rate
we extracted from a neonatal baby
we extracted from a neonatal baby
31
108892
5617
02:06
from a video we took
with a regular DSLR camera,
with a regular DSLR camera,
32
114509
2881
02:09
and the heart rate measurement we get
33
117390
1816
02:11
is as accurate as the one you'd get
with a standard monitor in a hospital.
with a standard monitor in a hospital.
34
119206
4811
02:16
And it doesn't even have to be
a video we recorded.
a video we recorded.
35
124017
2642
02:18
We can do it essentially
with other videos as well.
with other videos as well.
36
126659
2995
02:21
So I just took a short clip
from "Batman Begins" here
from "Batman Begins" here
37
129654
3901
02:25
just to show Christian Bale's pulse.
38
133555
1904
02:27
(Laughter)
39
135459
1822
02:29
And you know, presumably
he's wearing makeup,
he's wearing makeup,
40
137281
2123
02:31
the lighting here is kind of challenging,
41
139404
1953
02:33
but still, just from the video,
we're able to extract his pulse
we're able to extract his pulse
42
141357
2951
02:36
and show it quite well.
43
144308
2018
02:38
So how do we do all that?
44
146326
1920
02:40
We basically analyze the changes
in the light that are recorded
in the light that are recorded
45
148246
4598
02:44
at every pixel in the video over time,
46
152844
2271
02:47
and then we crank up those changes.
47
155115
1798
02:48
We make them bigger
so that we can see them.
so that we can see them.
48
156913
2162
02:51
The tricky part is that those signals,
49
159075
1902
02:52
those changes that we're after,
are extremely subtle,
are extremely subtle,
50
160977
2933
02:55
so we have to be very careful
when you try to separate them
when you try to separate them
51
163910
2779
02:58
from noise that always exists in videos.
52
166689
3831
03:02
So we use some clever
image processing techniques
image processing techniques
53
170520
2995
03:05
to get a very accurate measurement
of the color at each pixel in the video,
of the color at each pixel in the video,
54
173515
3994
03:09
and then the way the color
changes over time,
changes over time,
55
177509
2670
03:12
and then we amplify those changes.
56
180179
2693
03:14
We make them bigger to create those types
of enhanced videos, or magnified videos,
of enhanced videos, or magnified videos,
57
182872
3980
03:18
that actually show us those changes.
58
186852
2172
03:21
But it turns out we can do that
not just to show tiny changes in color,
not just to show tiny changes in color,
59
189024
4238
03:25
but also tiny motions,
60
193262
2241
03:27
and that's because the light
that gets recorded in our cameras
that gets recorded in our cameras
61
195503
3576
03:31
will change not only if the color
of the object changes,
of the object changes,
62
199079
2810
03:33
but also if the object moves.
63
201889
2368
03:36
So this is my daughter
when she was about two months old.
when she was about two months old.
64
204257
3636
03:39
It's a video I recorded
about three years ago.
about three years ago.
65
207893
2999
03:42
And as new parents, we all want
to make sure our babies are healthy,
to make sure our babies are healthy,
66
210892
3208
03:46
that they're breathing,
that they're alive, of course.
that they're alive, of course.
67
214100
2542
03:48
So I too got one of those baby monitors
68
216642
2142
03:50
so that I could see my daughter
when she was asleep.
when she was asleep.
69
218784
2469
03:53
And this is pretty much what you'll see
with a standard baby monitor.
with a standard baby monitor.
70
221253
3527
03:56
You can see the baby's sleeping, but
there's not too much information there.
there's not too much information there.
71
224780
3682
04:00
There's not too much we can see.
72
228474
1604
04:02
Wouldn't it be better,
or more informative, or more useful,
or more informative, or more useful,
73
230078
2824
04:04
if instead we could look
at the view like this.
at the view like this.
74
232902
2990
04:07
So here I took the motions
and I magnified them 30 times,
and I magnified them 30 times,
75
235892
6356
04:14
and then I could clearly see that my
daughter was indeed alive and breathing.
daughter was indeed alive and breathing.
76
242248
3826
04:18
(Laughter)
77
246074
2253
04:20
Here is a side-by-side comparison.
78
248327
1922
04:22
So again, in the source video,
in the original video,
in the original video,
79
250249
2483
04:24
there's not too much we can see,
80
252732
1636
04:26
but once we magnify the motions,
the breathing becomes much more visible.
the breathing becomes much more visible.
81
254368
3707
04:30
And it turns out, there's
a lot of phenomena
a lot of phenomena
82
258075
2070
04:32
we can reveal and magnify
with our new motion microscope.
with our new motion microscope.
83
260145
3623
04:35
We can see how our veins and arteries
are pulsing in our bodies.
are pulsing in our bodies.
84
263768
4564
04:40
We can see that our eyes
are constantly moving
are constantly moving
85
268332
2628
04:42
in this wobbly motion.
86
270960
1887
04:44
And that's actually my eye,
87
272847
1509
04:46
and again this video was taken
right after my daughter was born,
right after my daughter was born,
88
274356
3065
04:49
so you can see I wasn't getting
too much sleep. (Laughter)
too much sleep. (Laughter)
89
277421
4202
04:53
Even when a person is sitting still,
90
281623
2716
04:56
there's a lot of information
we can extract
we can extract
91
284339
2044
04:58
about their breathing patterns,
small facial expressions.
small facial expressions.
92
286383
3529
05:01
Maybe we could use those motions
93
289912
1625
05:03
to tell us something about
our thoughts or our emotions.
our thoughts or our emotions.
94
291537
3154
05:06
We can also magnify small
mechanical movements,
mechanical movements,
95
294691
3255
05:09
like vibrations in engines,
96
297946
1555
05:11
that can help engineers detect
and diagnose machinery problems,
and diagnose machinery problems,
97
299501
3692
05:15
or see how our buildings and structures
sway in the wind and react to forces.
sway in the wind and react to forces.
98
303193
4738
05:19
Those are all things that our society
knows how to measure in various ways,
knows how to measure in various ways,
99
307931
4581
05:24
but measuring those motions is one thing,
100
312512
2453
05:26
and actually seeing those
motions as they happen
motions as they happen
101
314965
2276
05:29
is a whole different thing.
102
317241
2554
05:31
And ever since we discovered
this new technology,
this new technology,
103
319795
3041
05:34
we made our code available online so that
others could use and experiment with it.
others could use and experiment with it.
104
322836
3953
05:38
It's very simple to use.
105
326789
1875
05:40
It can work on your own videos.
106
328664
2044
05:42
Our collaborators at Quanta Research
even created this nice website
even created this nice website
107
330708
3193
05:45
where you can upload your videos
and process them online,
and process them online,
108
333901
2678
05:48
so even if you don't have any experience
in computer science or programming,
in computer science or programming,
109
336579
3816
05:52
you can still very easily experiment
with this new microscope.
with this new microscope.
110
340395
2936
05:55
And I'd like to show you
just a couple of examples
just a couple of examples
111
343331
2404
05:57
of what others have done with it.
112
345735
2735
06:00
So this video was made by
a YouTube user called Tamez85.
a YouTube user called Tamez85.
113
348470
5317
06:05
I don't know who that user is,
114
353787
1463
06:07
but he, or she, used our code
115
355250
2345
06:09
to magnify small belly
movements during pregnancy.
movements during pregnancy.
116
357595
3715
06:13
It's kind of creepy.
117
361310
1602
06:14
(Laughter)
118
362912
1613
06:16
People have used it to magnify
pulsing veins in their hands.
pulsing veins in their hands.
119
364525
4961
06:21
And you know it's not real science
unless you use guinea pigs,
unless you use guinea pigs,
120
369486
3782
06:25
and apparently this guinea pig
is called Tiffany,
is called Tiffany,
121
373268
3390
06:28
and this YouTube user claims
it is the first rodent on Earth
it is the first rodent on Earth
122
376658
2949
06:31
that was motion-magnified.
123
379607
2688
06:34
You can also do some art with it.
124
382295
2188
06:36
So this video was sent to me
by a design student at Yale.
by a design student at Yale.
125
384483
3018
06:39
She wanted to see
if there's any difference
if there's any difference
126
387501
2137
06:41
in the way her classmates move.
127
389638
1522
06:43
She made them all stand still,
and then magnified their motions.
and then magnified their motions.
128
391160
4209
06:47
It's like seeing
still pictures come to life.
still pictures come to life.
129
395369
3378
06:50
And the nice thing with
all those examples
all those examples
130
398747
2433
06:53
is that we had nothing to do with them.
131
401180
2296
06:55
We just provided this new tool,
a new way to look at the world,
a new way to look at the world,
132
403476
3854
06:59
and then people find other interesting,
new and creative ways of using it.
new and creative ways of using it.
133
407330
5132
07:04
But we didn't stop there.
134
412462
1764
07:06
This tool not only allows us
to look at the world in a new way,
to look at the world in a new way,
135
414226
3251
07:09
it also redefines what we can do
136
417477
2368
07:11
and pushes the limits of what
we can do with our cameras.
we can do with our cameras.
137
419845
3181
07:15
So as scientists, we started wondering,
138
423026
2229
07:17
what other types of physical phenomena
produce tiny motions
produce tiny motions
139
425255
3785
07:21
that we could now use
our cameras to measure?
our cameras to measure?
140
429040
2903
07:23
And one such phenomenon
that we focused on recently is sound.
that we focused on recently is sound.
141
431943
4001
07:27
Sound, as we all know,
is basically changes
is basically changes
142
435944
2105
07:30
in air pressure that
travel through the air.
travel through the air.
143
438049
2183
07:32
Those pressure waves hit objects
and they create small vibrations in them,
and they create small vibrations in them,
144
440232
3621
07:35
which is how we hear
and how we record sound.
and how we record sound.
145
443853
2532
07:38
But it turns out that sound
also produces visual motions.
also produces visual motions.
146
446385
3668
07:42
Those are motions
that are not visible to us
that are not visible to us
147
450053
2833
07:44
but are visible to a camera
with the right processing.
with the right processing.
148
452886
3001
07:47
So here are two examples.
149
455887
1573
07:49
This is me demonstrating
my great singing skills.
my great singing skills.
150
457460
2614
07:53
(Singing)
151
461064
1634
07:54
(Laughter)
152
462698
1436
07:56
And I took a high-speed video
of my throat while I was humming.
of my throat while I was humming.
153
464134
2986
07:59
Again, if you stare at that video,
154
467120
1764
08:00
there's not too much
you'll be able to see,
you'll be able to see,
155
468884
2076
08:02
but once we magnify the motions 100 times,
we can see all the motions and ripples
we can see all the motions and ripples
156
470960
4332
08:07
in the neck that are involved
in producing the sound.
in producing the sound.
157
475292
3274
08:10
That signal is there in that video.
158
478566
2740
08:13
We also know that singers
can break a wine glass
can break a wine glass
159
481306
2670
08:15
if they hit the correct note.
160
483976
1463
08:17
So here, we're going to play a note
161
485439
1765
08:19
that's in the resonance
frequency of that glass
frequency of that glass
162
487204
2526
08:21
through a loudspeaker that's next to it.
163
489730
2048
08:23
Once we play that note
and magnify the motions 250 times,
and magnify the motions 250 times,
164
491778
4419
08:28
we can very clearly see
how the glass vibrates
how the glass vibrates
165
496197
2338
08:30
and resonates in response to the sound.
166
498535
3570
08:34
It's not something you're used
to seeing every day.
to seeing every day.
167
502105
2420
08:36
But this made us think.
It gave us this crazy idea.
It gave us this crazy idea.
168
504525
3529
08:40
Can we actually invert this process
and recover sound from video
and recover sound from video
169
508054
5608
08:45
by analyzing the tiny vibrations
that sound waves create in objects,
that sound waves create in objects,
170
513662
4035
08:49
and essentially convert those
back into the sounds that produced them.
back into the sounds that produced them.
171
517697
4777
08:54
In this way, we can turn
everyday objects into microphones.
everyday objects into microphones.
172
522474
4457
08:58
So that's exactly what we did.
173
526931
2232
09:01
So here's an empty bag of chips
that was lying on a table,
that was lying on a table,
174
529163
2816
09:03
and we're going to turn that
bag of chips into a microphone
bag of chips into a microphone
175
531979
2825
09:06
by filming it with a video camera
176
534804
1591
09:08
and analyzing the tiny motions
that sound waves create in it.
that sound waves create in it.
177
536395
3228
09:11
So here's the sound
that we played in the room.
that we played in the room.
178
539623
2796
09:14
(Music: "Mary Had a Little Lamb")
179
542419
4434
09:22
And this is a high-speed video
we recorded of that bag of chips.
we recorded of that bag of chips.
180
550007
3025
09:25
Again it's playing.
181
553032
1274
09:26
There's no chance you'll be able
to see anything going on in that video
to see anything going on in that video
182
554306
3342
09:29
just by looking at it,
183
557648
1058
09:30
but here's the sound we were able
to recover just by analyzing
to recover just by analyzing
184
558706
2984
09:33
the tiny motions in that video.
185
561690
2183
09:35
(Music: "Mary Had a Little Lamb")
186
563873
2809
09:52
I call it -- Thank you.
187
580985
1486
09:54
(Applause)
188
582471
5225
10:01
I call it the visual microphone.
189
589878
2345
10:04
We actually extract audio signals
from video signals.
from video signals.
190
592223
3390
10:07
And just to give you a sense
of the scale of the motions here,
of the scale of the motions here,
191
595613
3181
10:10
a pretty loud sound will cause that bag
of chips to move less than a micrometer.
of chips to move less than a micrometer.
192
598799
5336
10:16
That's one thousandth of a millimeter.
193
604135
2739
10:18
That's how tiny the motions are
that we are now able to pull out
that we are now able to pull out
194
606874
3561
10:22
just by observing how light
bounces off objects
bounces off objects
195
610435
3243
10:25
and gets recorded by our cameras.
196
613678
2136
10:27
We can recover sounds
from other objects, like plants.
from other objects, like plants.
197
615814
3250
10:31
(Music: "Mary Had a Little Lamb")
198
619064
6316
10:39
And we can recover speech as well.
199
627214
1997
10:41
So here's a person speaking in a room.
200
629211
2577
10:43
Voice: Mary had a little lamb
whose fleece was white as snow,
whose fleece was white as snow,
201
631788
4203
10:47
and everywhere that Mary went,
that lamb was sure to go.
that lamb was sure to go.
202
635991
4230
10:52
Michael Rubinstein: And here's
that speech again recovered
that speech again recovered
203
640221
2759
10:54
just from this video
of that same bag of chips.
of that same bag of chips.
204
642980
3274
10:58
Voice: Mary had a little lamb
whose fleece was white as snow,
whose fleece was white as snow,
205
646254
4831
11:03
and everywhere that Mary went,
that lamb was sure to go.
that lamb was sure to go.
206
651085
4859
11:07
MR: We used "Mary Had a Little Lamb"
207
655944
2346
11:10
because those are said to be
the first words
the first words
208
658290
2123
11:12
that Thomas Edison spoke
into his phonograph in 1877.
into his phonograph in 1877.
209
660413
4161
11:16
It was one of the first sound
recording devices in history.
recording devices in history.
210
664574
3228
11:19
It basically directed the sounds
onto a diaphragm
onto a diaphragm
211
667802
3327
11:23
that vibrated a needle that essentially
engraved the sound on tinfoil
engraved the sound on tinfoil
212
671129
4079
11:27
that was wrapped around the cylinder.
213
675208
2275
11:29
Here's a demonstration of recording and
replaying sound with Edison's phonograph.
replaying sound with Edison's phonograph.
214
677483
5943
11:35
(Video) Voice: Testing,
testing, one two three.
testing, one two three.
215
683426
3020
11:38
Mary had a little lamb
whose fleece was white as snow,
whose fleece was white as snow,
216
686446
3413
11:41
and everywhere that Mary went,
the lamb was sure to go.
the lamb was sure to go.
217
689859
3669
11:45
Testing, testing, one two three.
218
693528
2740
11:48
Mary had a little lamb
whose fleece was white as snow,
whose fleece was white as snow,
219
696268
4156
11:52
and everywhere that Mary went,
the lamb was sure to go.
the lamb was sure to go.
220
700424
5224
11:57
MR: And now, 137 years later,
221
705648
4017
12:01
we're able to get sound
in pretty much similar quality
in pretty much similar quality
222
709665
4087
12:05
but by just watching objects
vibrate to sound with cameras,
vibrate to sound with cameras,
223
713752
4079
12:09
and we can even do that when the camera
224
717831
1934
12:11
is 15 feet away from the object,
behind soundproof glass.
behind soundproof glass.
225
719765
4234
12:15
So this is the sound that we were
able to recover in that case.
able to recover in that case.
226
723999
3220
12:19
Voice: Mary had a little lamb
whose fleece was white as snow,
whose fleece was white as snow,
227
727219
5294
12:24
and everywhere that Mary went,
the lamb was sure to go.
the lamb was sure to go.
228
732513
4759
12:29
MR: And of course, surveillance is
the first application that comes to mind.
the first application that comes to mind.
229
737404
3630
12:33
(Laughter)
230
741034
2995
12:36
But it might actually be useful
for other things as well.
for other things as well.
231
744029
4056
12:40
Maybe in the future, we'll be able
to use it, for example,
to use it, for example,
232
748085
2840
12:42
to recover sound across space,
233
750925
2252
12:45
because sound can't travel
in space, but light can.
in space, but light can.
234
753177
3576
12:48
We've only just begun exploring
235
756753
2404
12:51
other possible uses
for this new technology.
for this new technology.
236
759157
3019
12:54
It lets us see physical processes
that we know are there
that we know are there
237
762176
2832
12:57
but that we've never been able
to see with our own eyes until now.
to see with our own eyes until now.
238
765008
3556
13:00
This is our team.
239
768564
1204
13:01
Everything I showed you today
is a result of a collaboration
is a result of a collaboration
240
769768
2879
13:04
with this great group
of people you see here,
of people you see here,
241
772647
2191
13:06
and I encourage you and welcome you
to check out our website,
to check out our website,
242
774838
3167
13:10
try it out yourself,
243
778005
1446
13:11
and join us in exploring
this world of tiny motions.
this world of tiny motions.
244
779451
2972
13:14
Thank you.
245
782423
1625
13:16
(Applause)
246
784048
1254
ABOUT THE SPEAKER
Michael Rubinstein - Research scientist, GoogleComputer scientist Michael Rubinstein and his team have developed a "motion microscope" that can show video footage of barely perceivable movements, like breaths and heartbeats.
Why you should listen
Michael Rubinstein zooms in on what we can't see and mangnifies it by thirty or a hundred times. His "motion microscope," developed at MIT with Microsoft and Quanta Research, picks up on subtle motion and color changes in videos and blows them up for the naked eye to see. The result: fun, cool, creepy videos.
Rubinstein is a research scientist at a new Cambridge-based Google lab for computer vision research. He has a PhD in computer science and electrical engineering from MIT.
Michael Rubinstein | Speaker | TED.com