Loading…

Am I the only one who sees a huge difference in BCI accuracy between 2019 and now? | brain-computer interface - Discusd

Am I the only one who sees a huge difference in BCI accuracy between 2019 and now?

I started messing with open-source BCI headsets back in 2019, and getting a steady signal was like pulling teeth. I'd spend 20 minutes just calibrating to detect a simple blink, and it would still drop out half the time. Fast forward to last month, I tried a newer dry electrode cap from a friend's lab, and it locked onto my alpha waves in under 2 minutes flat. The difference is night and day, but what actually changed in the hardware or software to make this happen? Has anyone else noticed this jump?

2 comments

2 Comments

ryang653mo ago

I saw a paper a few months back that basically confirmed what you're noticing. The big hardware change is in the front-end amplifiers, a lot of newer chips have way better common-mode rejection, so they cut down on power line hum and muscle noise before the software even gets the signal. But yeah, Christopher's right too, the software side is doing a lot of the heavy lifting now. I read somewhere that new adaptive filtering algorithms can basically learn the shape of your particular noise pattern in real-time and subtract it out, which is something the old fixed filters never could do. So it's kind of a one-two punch, better raw data coming in and smarter math cleaning it up faster.

christopher_flores463mo ago

You mentioned it locking onto alpha waves in under 2 minutes, but that's more about the software's ability to filter noise than the hardware being that much better. The big jump came from improved signal processing algorithms that can separate brain signals from muscle artifacts way faster than the old methods. Dry electrodes still have higher impedance than wet ones, so it's really the software doing the heavy lifting now.