Exploration of Human Emotion Based Real-time Memory and Computation Management on Resource-Limited Edge Devices
TimeWednesday, July 13th11:15am - 11:37am PDT
Location3004, Level 3
Embedded System Design Methodologies
DescriptionEmotional AI or Affective Computing has been projected to grow exponentially in the upcoming years. However, there has been a lack of hardware exploration leveraging the knowledge of users' emotions. In this paper, we establish a connection between users' emotions and the hardware and memory management of edge devices. Based on detection results from an efficient affect classifier, novel real-time management schemes on memory and video processing are proposed to improve the energy efficiency of mobile devices. Demonstrations on smartphone applications show significant energy saving or memory loading benefits when applying the proposed affect adaptive scheme to real-life usage scenarios.