Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Building an EEG with a Children's Toy (geofflord.substack.com)
84 points by gl44snip on Aug 17, 2024 | hide | past | favorite | 22 comments


Low-end EEG kits exist.[1] Probably easier to start with that.

My favorite toy in this space was Nekomimi in 2012 - cosplay cat ears connected to an EEG sensor.[2] It sensed "resting", "active", and "surprised". I saw some girls wearing these at a convention. Someone called out the name of one of them, and her ears popped up. Only the one, not the others.

It was a good idea, but too bulky. 4 AAA batteries plus huge ears. Someone should do that again, more compactly.

[1] https://www.amazon.com/NeuroSky-MindWave-Mobile-Brainwave-St...

[2] https://en.wikipedia.org/wiki/Neurowear



The MindWave Mobile seems far more convenient than a dissected Mindflex. Thanks for sharing the link—I may need to get one myself. Achieving consistent electrode placement was certainly a challenge when I was working on this project.

I hadn't come across the Nekomimi yet but it seems as though Neurosky has figured out how to manufacture relatively inexpensive EEG consumer devices. Further, the Mindwave mobile seems to get 8 hours of run time off a single AAA. I would suspect a more compact implementation of the Nekomimi is quite achievable.


And don't forget about the "Star Wars Science: Force Trainer"! (2009 - Uncle Milton" (toy company)


O hey, I have also done this before around 10 years ago. Initially my project was also based on a hacked Mindflex (it was a robot arm). I then reached out to NeuroSky and they just sent me a bunch of their higher tier dev kits for free. No idea if that’s still a thing, but I had really good experiences with them. The signals are a lot more useful.

I always found it funny that people playing Mindflex commonly believed the game was fake[1]. And it would have made sense - why go through the trouble of actually shipping an EEG sensor to children? But they did.

[1] Example: https://youtu.be/AJjN4se2yyQ


Good times! I also used the Mindflex hack[0] for an artsy piece in 2012. I attached it to an RF module (with a JeeNode[1]) and used the "attention" levels to control the playback speed of my hacky video player (zero attention -> 0.0x speed, full attention -> 2.0x speed). Worked great for the price!

[0]: https://frontiernerds.com/brain-hack (glad it's still online)

[1]: https://www.digitalsmarties.net/products/jeenode


Yes, that was it! I used the same hack in 2012. Fun story: as a prize for winning the science fair with my robot arm, I got to go to Munich to visit ASDEX Upgrade (fusion experiment). I was staying there for a week and just spent each day with a different department.

They have a post-lunch lightning talk thing where interested people gather in a lecture hall with a cup of coffee to listen to a random talk. I was asked to demonstrate my project last-minute but did not have time to ship all the stuff to where I was. I had the headset though (an upgraded NeuroSky model). So I spent the night hacking away at a demo project that just involved kicking a virtual football (soccer) into a goal.

So at some point in 2012 there was a room full of nuclear scientists watching a teenager kick a virtual ball with "their mind". I just find that amusing.


That's awesome! Did you attach the electrodes to the head or somewhere on your arm? I would also be very interested to know how you decoded the electrode data.

On another note, I found it pretty surprising that it's possible to get even remotely accurate data from the Mindflex. As a child, I remember wholeheartedly believing it was real. But, only when I won.


I will also need to shoot NeuroSky a message to see if I can get my hands on one of their dev kits. This recent venture into decoding EEG data was fascinating and I would love to see how far I could take it with non-invasive electrodes.


I also recommend this video as a resource on low cost EEG from someone who tried living with it: https://youtu.be/c-LyXBCNsrA?si=7xoM1OGTNRbMsO_P (skim through)

That goes over several nicer multichannel options. The NeuroSky stuff will only get you so far :)


This is great, thanks for sharing. I think I will give a multichannel headset a try and attempt to train some models that map neural activity to cursor movements.

I came across this article today: https://medium.com/@kevinjycui/of-mice-and-mind-creating-a-s...

It sounds as if this individual had some success with the Emotiv INSIGHT (5 Channel) but the data quality was a limiting factor in creating an application usable for everyday activity.


Wait don’t leave us hanging. What were your results?


I was in high school and this was for a science fair. It kind of worked, you could more or less reliably move one axis of the arm at a time. I directly used the relax/focus metrics and relied on a training phase where the participant would try to find thoughts/states of mind that would move the needle the most. Very basic stuff (but I didn’t know better at the time).


Fantastic project IMHO


I am reminded of the meme: Thinking quickly, Dave constructs a homemade [EEG], using only some string, a squirrel, and a[n EEG]. That said, breaking out the functionality from the toy is neat+useful.


This gave me a good laugh! Perhaps a better title would have been "Tearing Apart a Toy to Obtain EEG Data". Regardless, I am glad you found it interesting.


I mean, getting an EEG for ~$20 (current price on eBay) and figuring out how to use it is an impressive feat in its own right!


I use the Muse EEG band [1] but with the Mind Monitor third party app and tools [2]. This is another case of an official app being subpar to the hardware device.

[1] https://choosemuse.com/?srsltid=AfmBOopzyEsQF8LcBiGJD4U9fKRR...

[2] https://mind-monitor.com/


Did you have any specific gripes with the Mind Monitor app? The visualizations seem nice, but it would be interesting to see a feature that allows synchronization with other data sources. (e.g. connecting a keyboard to capture both neural activity and keystrokes simultaneously) That said, this might not be worth the effort. For me, obtaining high-quality data was a challenge, which I suspect is a common limitation of non-invasive electrodes.


You can export your data and work with it. Also, there are some online tools and docs [1][2]. Also, there are open source libs such as [3]. My only "gripe" is knowing enough about EEG.

[1] https://mind-monitor.com/Chart.php

[2] https://mind-monitor.com/Macro.php

[3] https://www.npmjs.com/package/muse-js


Me: wow what a cool cutting edge AI project idea.

Article: yeah just dusting off an old toy I used as a kid

HN: $various relevant stories from 10+ years ago

Me: need to leave cave

Cool project

Looks like autocorrect has turned the learned gentlemans perceptrons into perceptions.


It seems as if the Overleaf autocorrect has a mind of its own occasionally! Or, that's my perception at least.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: