Ambient Temperature Estimation using Neural Networks and Device Contextual Information
TimeWednesday, July 13th6pm - 7pm PDT
LocationLevel 2 Lobby
Event Type
Networking Reception
Work-in-Progress Poster
DescriptionAmbient temperature impacts the smartphone device temperature, making its knowledge crucial to design efficient thermal management policies. However, use of on-board ambient temperature sensors is less because their measurements are influenced by heat generated from various hardware components. This paper focuses on an on-device ambient temperature estimation model that provides the ambient temperature without using dedicated ambient temperature sensors. With an average inference time of 20ms, the proposed model can estimate ambient temperature with an error margin of 1°C for 99.6% points. Application of this research can improve a smartphone’s thermal awareness and improve user experience.