Configuration folding: An energy efficient technique for context recognition

摘要

This demonstration presents configuration folding, a novel technique which enables the energy efficient execution of multiple context recognition applications on a personal mobile device. Configuration folding achieves energy efficiency by identifying and removing the redundant functionalities in different context recognition applications at runtime. The demonstration shows the effectiveness of configuration folding. To do this, we have implemented two non-trivial sound-based context recognition applications that are able to detect speech and music. During the demonstration, we will execute these applications with folded and unfolded configurations and we will show how the resulting energy gains can be measured. Furthermore, we present how similar applications can be built by means of composition.