Recently, active matrix organic light-emitting diode (AMOLED) displays have widely been used for user-interface applications on smartphones. Since frequent interactions between a user and a display which come from such applications tend to incur big display power consumption, the display power efficiency should be managed exactly and fast. For this purpose, a high-accurate and fast power model for AMOLED displays is greatly required for efficient power management. In this paper, we propose a novel high-accurate and fast power model for AMOLED displays, which is developed using a multiple regression technique to account for the dependency between R, G, and B channels and is polished using various performance optimization techniques. Then, the proposed model is applied to five AMOLED display panels of a CHIMEI panel and Galaxy S1, 3, 5, and 7 panels from SAMSUMG. The average error rates are obtained on these panels for four well-known image datasets. The proposed model shows the lowest error rate of 1.29% on Galaxy S7, while the simple model and Park's model show the error rates of 6.91% and 6.57%, respectively. In addition, in order to guarantee the fast execution time of the proposed power model on smartphones, we implement performance optimizations to the model. For runtime application, the proposed power model is implemented at the operating system level of the Android smartphone equipped with Exynos 4412. The optimized power model shows an average execution time of approximately 7.77 milli-seconds per image displayed on the panel with 720x1280 resolution. Consequently, the proposed power model is shown to be very fast and accurate, and thus it is highly useful in smartphones with AMOLED displays.
This work was supported by the National Research Foundation of Korea (NRF) Grant funded by the Korean Government (MSIT) under Grant 2018R1A2B6005466 and in part by the Ajou University Research Fund.
