Title: Emotion Recognition from Vocal Streams Captured with Microsoft Hololens Date: 2018-10-01 19:30 Modified: 2010-12-05 19:30 Category: research Tags: img: https://github.com/AtomScott/emotion-recognition-from-voice/blob/master/img.jpg?raw=true link: https://github.com/AtomScott/emotion-recognition-from-voice Slug: emotion-recognition-from-voice Authors: Atom Scott @ Xperception Lab, University of Tsukuba Summary:
This research stemmed from an R&D project with an undisclosed Japanese research center.
※More info may be released in the future.
Thanks to my previous experiance in modelling continous time-series data, I was assigned to implement a module that can do emotion recognition from vocal streams captured with the Microsoft Hololens.
I used a naive deep-learning based approach and although the results where sub-optimal, precision was much better than random data and was acceptable for this R&D project.
However, I wanted to improve the results and learn more about audio processing, and so I have decided to continue to do some research in the field. I plan to release a paper with my findings later this year.