<s>
Skinput	B-Device
is	O
an	O
input	B-General_Concept
technology	I-General_Concept
that	O
uses	O
bio-acoustic	O
sensing	O
to	O
localize	O
finger	O
taps	O
on	O
the	O
skin	O
.	O
</s>
<s>
When	O
augmented	O
with	O
a	O
pico-projector	B-General_Concept
,	O
the	O
device	O
can	O
provide	O
a	O
direct	O
manipulation	O
,	O
graphical	B-Application
user	I-Application
interface	I-Application
on	O
the	O
body	O
.	O
</s>
<s>
Skinput	B-Device
represents	O
one	O
way	O
to	O
decouple	O
input	O
from	O
electronic	O
devices	O
with	O
the	O
aim	O
of	O
allowing	O
devices	O
to	O
become	O
smaller	O
without	O
simultaneously	O
shrinking	O
the	O
surface	O
area	O
on	O
which	O
input	O
can	O
be	O
performed	O
.	O
</s>
<s>
While	O
other	O
systems	O
,	O
like	O
SixthSense	B-General_Concept
have	O
attempted	O
this	O
with	O
computer	O
vision	O
,	O
Skinput	B-Device
employs	O
acoustics	O
,	O
which	O
take	O
advantage	O
of	O
the	O
human	O
body	O
's	O
natural	O
sound	O
conductive	O
properties	O
(	O
e.g.	O
,	O
bone	O
conduction	O
)	O
.	O
</s>
<s>
Skinput	B-Device
has	O
been	O
publicly	O
demonstrated	O
as	O
an	O
armband	O
,	O
which	O
sits	O
on	O
the	O
biceps	O
.	O
</s>
<s>
Classification	O
is	O
driven	O
by	O
a	O
support	B-Algorithm
vector	I-Algorithm
machine	I-Algorithm
using	O
a	O
series	O
of	O
time-independent	O
acoustic	O
features	O
that	O
act	O
like	O
a	O
fingerprint	O
.	O
</s>
<s>
Like	O
speech	B-Application
recognition	I-Application
systems	O
,	O
the	O
Skinput	B-Device
recognition	O
engine	O
must	O
be	O
trained	O
on	O
the	O
"	O
sound	O
"	O
of	O
each	O
input	O
location	O
before	O
use	O
.	O
</s>
<s>
With	O
the	O
addition	O
of	O
a	O
pico-projector	B-General_Concept
to	O
the	O
armband	O
,	O
Skinput	B-Device
allows	O
users	O
to	O
interact	O
with	O
a	O
graphical	B-Application
user	I-Application
interface	I-Application
displayed	O
directly	O
on	O
the	O
skin	O
.	O
</s>
<s>
Despite	O
being	O
a	O
Microsoft	O
Research	O
internal	O
project	O
,	O
Skinput	B-Device
has	O
been	O
demonstrated	O
publicly	O
several	O
times	O
.	O
</s>
<s>
The	O
first	O
public	O
appearance	O
was	O
at	O
Microsoft	O
's	O
TechFest	O
2010	O
,	O
where	O
the	O
recognition	O
model	O
was	O
trained	O
live	O
on	O
stage	O
,	O
during	O
the	O
presentation	O
,	O
followed	O
by	O
an	O
interactive	O
walkthrough	O
of	O
a	O
simple	O
mobile	O
application	O
with	O
four	O
modes	O
:	O
music	O
player	O
,	O
email	O
inbox	O
,	O
Tetris	B-Device
,	O
and	O
voice	O
mail	O
.	O
</s>
