Backyard Brains Logo

Neuroscience for Everyone!

+1 (855) GET-SPIKES (855-438-7745)

items ()

Pick a Card Project Update: Deploying Hardware for TinyML Model

Pick a Card Project Update

— Written by Nour Chahine —

Before I start discussing my updates, I have an announcement to make: I solemnly swear that I will solely be working with SSVEPs for the remainder of my project!

Over the past few weeks, I worked on feeding different forms of data into the neural network. I mainly applied two approaches: the power spectral density vector method and the notch filter knockout vector method. Both methods rely on quantifying the EEG response relative to the flashing frequencies of the cards.

Power Spectral Density Vector

The power spectral density vector method involves computing the power spectral density of the EEG signal, and then making a vector of the total power of the signal at each of the target frequencies.


Don’t Blink, You Might Miss It! Can Machine Learning Be Our Eyes and Ears?

Can Machine Learning Be Our Eyes - hardware for FOMO glasses

— Written by Ariyana Miri —

Welcome back FOMO gang!

If you’ve been following along with my FOMO glasses journey, then you know I’m trying to build a pair of glasses that capture a photo every time you blink. In my first post, we discussed the idea behind my project, and the implications the success of the project can have. In the second, we discussed the nature of an eye blink signal, and how the data is being processed. Today, we’re talking about hardware and how I’m actually putting the thing together, so check out the other posts if you haven’t had the chance yet!

Can Machine Learning Be Our Eyes - layered blinks

With my project, while the signal of an eye blink is stronger and easier to detect than others, the hardware aspect presents an entirely new challenge. How are we going to get everything to communicate together and fit on a pair of glasses?


Bluff or No Bluff: Can a Neural Network Tell Them Apart?

Bluff or No Bluff - Sachin playing poker online

— Written by Sachin Pillai —

Hello to everyone out there! This is Sachin here again with updates on your very own portable poker bluff detector. I ended my last blog post mentioning how I developed the capability to collect electrodermal activity during full rounds of poker. This has allowed me to amass a large dataset of subjects playing poker. I am now nearing the end of my data collection/analysis phase and am close to producing a functional bluff detection method.

The past three weeks have consisted of lots of dead-ends, retracing, and fine-tuning. My initial approach to classifying bluff from neutral signals was to use the raw images of the galvanic skin responses and feed these graphs into the neural network. Just by judging these samples with my eye, I was able to somewhat distinguish between samples that contained a bluff versus those that did not. Data samples representing these two scenarios are shown below.

the difference in galvanic skin response 1
the difference in galvanic skin response 2

Both images represent electrodermal activity during a live poker session. Each image contains the GSR (galvanic skin response) and pulse response five seconds after either the flop, river, or turn has been uncovered. The image on the left is data collected during a neutral trial where the player was not disadvantaged by the cards on the table and did not decide to bluff. The image on the right shows when a player made a conscious decision to bluff after seeing the cards on the table. The orange line (SBChannel1) is the galvanic skin response while the green line (SB Channel 2) represents the pulse rate.