Energy Measurement and Optimization of Continuous Gesture Recognition

Student thesis: Master thesis (including HD thesis)

  • Jens Emil Gydesen
4. term, Computer Science, Master (Master Programme)
Energy consumption of software is becoming increasingly important as wearables, such as fitness trackers and other small battery-powered devices, are becoming a large part of our lives. These devices are typically controlled by a low-power CPU and powered by a small battery, and are thus very resource constrained. This thesis proposes a simple yet accurate method for developers to measure the energy consumption of their software. We demonstrate this method by implementing two continuous gesture recognition algorithms, and measuring their energy consumption. Furthermore we propose optimizations for these algorithms and gain up to 22 % energy reduction with no loss of accuracy, and up to 92.36 % energy reduction with 11 % loss of accuracy.
Publication date7 Jun 2016
Number of pages37
ID: 234925168