Aurora Biometrics

Inspiration

We at Aurora are aspiring quant traders, quant developers, and quant researchers. A current fun group effort of ours entails us dedicating the majority of this month towards learning the mathematics behind poker, our initial plan was to build an efficient black jack bot and spend the rest of our time networking with fellow quant finance aspirants and playing poker, till we struck the idea that by using a dataset, we can turn human behaviorisms into a quantifiable metric and leverage it in games like poker! Now here we are! We built the system that estimates a participants uncertainty level using real world data collected here at TAMU.

What it does

Our device tracks biometric data from human faces using OpenFace, a Python-based facial recognition framework, to monitor subtle physiological responses in real time. By assessing action units, it calculates a continuous stress and baseline deviation score for each participant, allowing us to model that to the user on a web-app.

How we built it

We captured high-resolution video using an iPhone setup while guiding Texas A&M students through a structured sequence of baseline and high-stakes questions. The extracted video frames were processed via the OpenFace CLI to compute facial features, which were then normalized and evaluated using a cross-validated Random Forest machine learning pipeline.

Challenges we ran into

Standardizing baseline facial dynamics across multiple individuals proved difficult due to varying natural expressions. Additionally a low sample size as a result of a low number of willing participants also proved difficult.

Accomplishments that we're proud of

We successfully implemented a framework that evaluates new participants real time against model parameters without prior exposure to their faces. Furthermore, we translated complex frame-by-frame action unit shifts into a uncertainty metric .

What we learned

We gained deep insights into feature engineering around facial action units and how empirical weighting can magnify subtle physiological leakage indicators like eyebrow asymmetry. We also gained skills in conducting a survey across a population.