Automated Inference of Cognitive Stress in-the-Wild

Abstract

We aim to build technology that combines mobile
sensing systems to automatically infer a person's
cognitive stress to provide better and continuous stress
management support. Our main innovation is the use
of low-cost mobile thermal camera integrated in
smartphone or other devices to produce new stress
measures. We have developed a robust mobile based
tracking system that tracks a person's breathing
pattern by measuring temperature changes around a
person's nostrils region while the person is facing the
smartphone. Stress levels are automatically
assessed by capturing breathing pattern dynamics
through a novel signature based on time and frequency
values and using convolutional neural
networks to reduce the need to hand craft
higher level features. We are now extending the system
to integrate multiple sensors (e.g., PPG and GSR) and
behavioural information (context). The system is being
also adapted to be applied in the context of industry
workfloor within the EU H2020 HUMAN research project to support
workers during stress inducing tasks. Evaluations are
being conducted both in the laboratory and in-the-wild
(e.g., industry workfloor).