Datasets:
olimiemma
/
MIT-OCW-Transcripts

Dataset card Data Studio Files
xet
 Community
1
text
stringlengths
0
205
I know British
sign language too,
but that's another
sign language.
PROFESSOR: So you mentioned
when you're in a party
and you can't hear people,
does that mean that there's
a lot of noise that
masks speakers and that's
a hard situation for you?
Right.
SHEILA: So like
the speaker's voice
will blend into other speakers
voices or background noises,
so I tend to rely on
lipreading or some other method
to communicate.
PROFESSOR: Right.
So for example, in
cochlear implants,
a common problem is when
there is an environment where
there's many, many
frequencies of sound,
like a crowded
restaurant or a party,
and there's one speaker that
you're trying to pay attention
to and the subject
gets overloaded
on every single electrode.
And so some kinds of
cochlear implant processors
try to circumvent that by trying
to pick out in the spectrum
the important peaks
of the spectrum.
So if you're listening
to the vowel aa,
you'd have three formants.
The processor tries to pick
out those formants and only
present electrodes
corresponding to those formants
and turn all the other
electrodes off so that there's
a huge difference between
where the formant is
and where the nothing is.
Really in theory, it's
nothing, but actually, it
could be a noisy background.
So that is one kind of
speech processor design.
It's called the speech
feature extractor, sometimes
the speak chip.
It's trying to pick
out formants so
that it can understand vowels.
And it's supposed to be less
sensitive to noise masking,