Thanks to the wonderful Celia Hodent and her guidance, I was accepted for a talk on my PhD topic for the UX Summit. It was quite surreal experience. Once that I will never forget. While the talk can be accessed via the GDC Vault, it is currently only available to members. As a “2022-is-ending-treat” I would like to share my slides with you.
Feel free to contact me anytime. I am currently taking a break from the PhD project until 2024, however, I am still looking for potential collaborators (EDA analytics).
@Betty Chacko
Hi Betty,
Thanks for reaching out, and happy to help where I can.
EDA is a broad area, so it would help to know where you’re starting from. The main things I’d flag up front:
1. EDA indexes arousal, not valence. On its own it can’t separate fear from excitement. For actual emotion classification you generally need it fused with another signal (facial EMG, HR/HRV) or with strong contextual constraints.
2. Decomposition into tonic (SCL) and phasic (SCR) components does most of the work. cvxEDA or CDA (Ledalab) are the usual choices; classifying on the raw signal rarely works well.
3. Normalize within subject. Absolute microsiemens values aren’t comparable across people, so range-correction or within-subject z-scoring isn’t optional.
If you let me know what you’re working with I can give you something more specific. Happy to point you to the relevant literature too.
Best,
Natt
Hi Natali,
Greetings!
May I get some clarifications on electro dermal activity signals’ (EDA) analysis for emotion classificaton?
My name is Betsy, Research scholar and Asst Professor in collegiate education, Kerala, India.
Thank you in Advance,
Betsy.