The Goosecam
Experiment 1: Detecting emotions from wearable devices
Timeline
February 2018 (Founding Date) — July 2018
But Why?
For the past few months, I had been working on detecting human emotions from brain waves (EEG) in a neuroengineering lab at UCLA.
I love music and travel, and choice overload was becoming a really big hinderance: I was constantly spending more time choosing what to listen to or where to go next, rather than actually enjoying anything.
Everyone around me was obsessed with creating a better future; they were building devices that would enable blind people to see, paralysed people to walk, and make brain to brain communication a reality. On the contrary, I wanted to work on optimising our present moments; solving choice overload via wearable devices seemed like the perfect problem to attack.
Experiment
Lack of context, and real-time feedback are the primary reasons behind choice overload. When we go out with a friend, they can sense our context (how we are feeling, where we are at, who else are we with etc.) very intuitively, and react to our implicit feedback (tapping of fingers, singing along etc.) in real-time. To build our target recommendation system we basically need to recreate this friend.
This experiment was aimed at solving the context problem specifically. Everything except the emotion sensor is already available quite easily, so we focused on figuring out ways of detecting peak emotional responses to musical stimuli via wearable devices.
We assembled a team of experts in the field of psychophysiology, and tried several signals like EEG and EKG, and finally settled for piloerection (goosebumps). Whenever, we really enjoy something we get goosebumps, and if we can somehow capture them automatically in real-time we can recreate goosebumps on demand via musical stimulus.
We designed a device that would have the form of a wrist band, and contain a camera to detect goosebumps (goosecam), in addition to location, motion and weather sensors. These combined can recreate the contextual understanding humans have.
Observations
EEG, EKG etc. signals worked well in lab settings, but in real life settings with a lot of muscle movement, and sweat, the signals can be heavily distorted. Piloerection is immune to this noise as it relies on a camera.
The smaller the goosecam surface area, the lower the precision with which it could detect goosebumps. This would be a major roadblock in commercial hardware; users do not want a bulky device on their hands all the time.
Even the most ardent music lovers were very reluctant to buy a new wearable just for the sake of music recommendations.
Although the designing and prototyping cost for the hardware was minimal, producing it at commercial scale required large manufacturing orders, and costs.
Conclusion
People do not want one more wearable, they want one wearable that does everything, and it is stupid to compete with companies like Apple on this front.
Rather than trying to create really advanced wearable tech just because it’s a cooler and more direct solution, we should account for some noise, and try out approaches to detect context from smartphones directly.
We should switch from wearables to a mobile-only emotion/context detection system.