Toward a System for Longitudinal Emotion Sensing

Abstract

The ability to sense acoustic emotion from a smartphoneis advantageous for two main reasons. First, smartphonesensing is unobtrusive compared to wearing a microphone, and second, a smartphone is nearly always with the user. When sensing emotion over a long period of time, these two reasons become increasingly more important. We demonstrate the challenges of building a system for longitudinal emotion sensing on a smartphone, as well as our design approach to these challenges. Current emotion sensing systems perform all sensing and computation on the phone, but this design can lead to significant battery life constraints. We show that in terms of energy consumption, offloading feature and classification computation to a remote server is the most feasible design choice without excessive battery draining, and discuss how we address the privacy, energy, and storage concerns of such an approach.