About

Independent research using hardware loaned by the GFT Labs Digital Innovation Lab. What I built, what I found, and what I would hand to the next person who picks one up.

The Neurosity Console live view, where this research happened
219

trials audited

248,320

samples

8

sensors

256 Hz

sampling rate

3

sessions read

Samantha Lin

Who made this

I'm Samantha Lin. I'm a Master's student in Human Computer Interaction at Elisava School of Design and Engineering, and I worked with the GFT Labs Digital Innovation Lab, who loaned me a Neurosity Crown so I could find out what it does.

My first degree was a Bachelor's in Psychology with a minor in Sociology, and the research training from it turned out to matter more for building on the Crown than anything in the EEG textbooks: what the literature actually supports, and where the honest line sits between a published finding and my own reasoning. That habit runs through everything here. Where a claim is mine rather than the literature's, I say so.

This is independent research using hardware loaned by the GFT Labs Digital Innovation Lab, and it feeds my own thesis project. My part was hands-on and self-directed. I set the training protocol, ran the sessions, audited what came back, and wrote the tools that read it.

The pivot away from training the headset was not a quick decision. I had 219 trials recorded and a protocol I still think was right, and the honest reading of the situation was that continuing would produce a model only I could use, built on trials I could never open and check. Stopping cost me the goal I had started with. What it bought was every finding on this site, which is the part that transfers to anyone else who picks up this hardware. A logger, crown-focus-logger, records sessions to CSV you own. A debrief, crown-debrief, turns a recording into plain English with no AI model in the loop. A built-in guide answers questions about a session and cites its sources. Alongside the tools I wrote down everything I learned that is not in the documentation.

What that came to: 219 training trials across 17 experiments, three recording sessions read closely, 248,320 samples of raw voltage, and two tools plus a guide for turning a session into something a person can read. The whole arc, including the wall I hit and the pivot it forced, is in the case study.

More of my work: samanthalin130@gmail.com and LinkedIn.

Where this came from

My Master's thesis project is AuraPod AI. The Crown entered that work as the way to measure my own neurophysiological state accurately enough to pair with it, rather than as a subject in its own right.

This site is the loaned-hardware research that grew out of that pairing. Once I started trying to read the signal properly, the reading turned out to be the harder and more useful problem.

The AuraPod AI exhibition setup at a graduate show, with the Neurosity Crown headset on display as part of it.
My AuraPod AI exhibition setup, with the Crown as the measurement layer.

What a Neurosity Crown is

The Crown is a consumer EEG headset, a Crown 3 in this case: eight dry recording electrodes resting against the scalp, sampling at 256 times a second. Dry means no gel and no preparation, which makes it quick to put on and more sensitive to how well it is sitting. It is not a medical device, and nothing on this site is a health assessment.

Two things come out of it. The raw signal, eight channels of microvolts, which you can record and export yourself. And two scores, focus and calm, which are not measurements: they are the output of models trained on other people, delivered as a probability between 0 and 1.

It also ships with a training feature. You pick a command like "left hand pinch," record yourself imagining it thirty times, and a model on Neurosity's servers learns to spot that pattern in your live signal. That feature is where this project started, and where it stalled.

Official Neurosity documentation covers the setup, the API, and the numbers, at docs.neurosity.co. This site covers what the documentation doesn't: the console screen by screen, which is also where the vendor's own figures are kept, and the things I wish someone had handed me first.

The Neurosity Crown, a matte black headband with a curved band across the forehead and arms reaching over the head.
Photo: Neurosity media kit
Top-down map of the head showing eight data sensors, two reference-only sensors at T7 and T8, and a dashed midline with no sensor.
Eight recording sensors, and a gap down the middle with none.

Questions, answered

Decode the console
A Neurosity Crown, a consumer EEG headset with 8 recording electrodes sampling at 256 times a second. It ships with a training feature: you pick a command like "left hand pinch," record yourself imagining it thirty times, and a model on Neurosity's servers learns to spot that pattern in your live signal.
They're scores from models trained on other people, and they come out as a probability from 0 to 1. A focus of 0.42 means a model put the chance you were focused at 42%, not that it measured 42% of anything.
No. The console shows a count and that is all: no list of trials, no score per trial, no export button, nothing. If you want data you actually own, record it yourself with the SDK instead of training in the console.
Recorded sessions, and any notes written against them, stay on your own machine. The handoff version is designed as a set of static files with no server behind it: when someone drops a CSV into it, the file is read inside their own browser and goes nowhere. No upload, no database, nothing to breach.
No. This is a consumer headset, not a medical device. The app describes what the data did and stops. Nothing here is a health assessment.

Explore more

This site covers what the documentation doesn't: the things I wish someone had handed me first.