4 ms·

Anybody has experience and is it actually useful? Heard a lot of "it captures mostly face muscle contractions" and "capturing a brain signal with it is like li

Anybody has experience and is it actually useful? Heard a lot of "it captures mostly face muscle contractions" and "capturing a brain signal with it is like listening to a whisper being spoken in a grand canyon, while being miles away"

amelius 2y ago
> it captures mostly face muscle contractions Maybe it is useful for people who suffer from bruxism.
lkeskull 2y ago
Listening to a whisper in the grand canyon is possible with modern ML ;)
etrautmann 2y ago
As a neuroscientist focused on intracortical recordings, these can be useful for certain clinical diagnostics but they’re quite limited in terms of signal quality. They signals are analogous to listening to an orchestra from outside the building. You may be able to tell that a song is playing but you lose the vast majority of the detail, and what we can learn from this is somewhat limited. Noise from facial muscles is another huge issue if you’re focused on the brain. Some interface devices focus on these signals instead for silent speech decoding etc. that can work, but the applications are still somewhat niche.
LeifCarrotson 2y ago
It is a lot like that. I only played with these briefly in biomedical engineering, we experimented with signal analysis algorithms while monitoring engineering students on varying levels of caffeine (it was hard to find volunteers for the control group) and doing various tasks. You could correlate EEG measurements to tell whether someone was focusing on a reaction time test and exactly when they reacted, whether they were at a stressful or relaxed part of a movie, estimate their WPM level while typing (calibrated to individual users), and similar stuff, which was at least interesting. I think one group did a sleep study and correlated EEG readings to heart rate and EMG data, looking for sleep phases and REM state activity levels, which is a common use case too. Whether it's useful or not is a completely separate question, but those are the kinds of things that you can measure. Some people (eg. Youtuber Perrikaryal) are trying various things like mapping "concentrate really hard on pushing this cube" to moving a video game character on a screen, which is astonishingly impressive to me, whose last exposure to the field is over a decade old. But even at the bleeding edge, it's nowhere near the point of "I'm imagining a shape, now 3D print it" or anything like that. Just as with AI, the hype in this sector is high because people imagine that it will eventually do that and they want to be the first ones there, but there's a lot of work between then and now if it's even possible at all.

amelius 2y ago

> it captures mostly face muscle contractions Maybe it is useful for people who suffer from bruxism.

lkeskull 2y ago

Listening to a whisper in the grand canyon is possible with modern ML ;)

etrautmann 2y ago

As a neuroscientist focused on intracortical recordings, these can be useful for certain clinical diagnostics but they’re quite limited in terms of signal quality. They signals are analogous to listening to an orchestra from outside the building. You may be able to tell that a song is playing but you lose the vast majority of the detail, and what we can learn from this is somewhat limited. Noise from facial muscles is another huge issue if you’re focused on the brain. Some interface devices focus on these signals instead for silent speech decoding etc. that can work, but the applications are still somewhat niche.

LeifCarrotson 2y ago

It is a lot like that. I only played with these briefly in biomedical engineering, we experimented with signal analysis algorithms while monitoring engineering students on varying levels of caffeine (it was hard to find volunteers for the control group) and doing various tasks. You could correlate EEG measurements to tell whether someone was focusing on a reaction time test and exactly when they reacted, whether they were at a stressful or relaxed part of a movie, estimate their WPM level while typing (calibrated to individual users), and similar stuff, which was at least interesting. I think one group did a sleep study and correlated EEG readings to heart rate and EMG data, looking for sleep phases and REM state activity levels, which is a common use case too. Whether it's useful or not is a completely separate question, but those are the kinds of things that you can measure. Some people (eg. Youtuber Perrikaryal) are trying various things like mapping "concentrate really hard on pushing this cube" to moving a video game character on a screen, which is astonishingly impressive to me, whose last exposure to the field is over a decade old. But even at the bleeding edge, it's nowhere near the point of "I'm imagining a shape, now 3D print it" or anything like that. Just as with AI, the hype in this sector is high because people imagine that it will eventually do that and they want to be the first ones there, but there's a lot of work between then and now if it's even possible at all.