A Comprehensive Study of Performance-Autonomy Trade-off on Smart Connected Glasses

Embedded detection systems presented in the form of wearable devices become very useful in the healthcare area. However, such devices have strong limitations in terms of size, energy and computational resources, and must be therefore carefully designed. In this paper we propose to take into account the QoS during the power consumption study, so that both performance and autonomy can be conjointly optimized for smart connected glasses. To do so, a methodology based on a system level power modeling approach and a performance characterization using several performance metrics is introduced. As part of this performance characterization, a comprehensive study for finding the system configuration that offers the best tradeoff between performance and autonomy is proposed for a blink detection application.