Modeling power consumption for video decoding on mobile platform and its application to power-rate constrained streaming

Abstract

This paper proposes an analytical power consumption model for H.264/AVC video decoding using hardware (HW) accelerator on popular mobile platforms. Our proposed model is expressed as the product of the power functions of video spatial resolution (i.e., frame size) and temporal resolution (i.e., frame rate). We have demonstrated that the same analytical model is applicable to different platforms. Model parameters are fixed for a specific platform. This indicates that HW accelerated video decoding is independent of the video content. Simulation results show the high accuracy for video decoding power prediction using proposed model, with the maximum relative prediction error less than 10%. Together with the video bit rate and perceptual quality models published in separated works, we propose to solve the power-rate optimized mobile video streaming problem, so as to maximum the video quality given the limited access network bandwidth and battery life for mobile devices.

Power consumption and bandwidth savings with video transcoding to mobile device-specific spatial resolution

“…In the experiment we put the priority on lowering spatial resolution while keeping quantization parameter (QP) fixed to 24 for all video sequences and resolutions. The paper [4] showed that quantization parameter doesn't affect the power consumption unlike spatial resolution. The acceptability-based model also showed that PAcc drops faster while increasing QP than it drops while reducing spatial resolution.…”

“…It is important to minimize power consumption of video decoding and network requirements without drastically affecting QoE. Xin Li et al [4] proposed an analytical power consumption model for H.264/AVC video decoding using a hardware accelerator on popular mobile platforms. The model is expressed as a product of the power functions of video spatial resolution and temporal resolution.…”

A pixel-based complexity model to estimate energy consumption in video decoders

“…Using likwid-perfctrs marker API it is possible to measure the energy consumption of selected functions of an application by turning on/off hardware performance counters, which allows to obtain the CPU energy consumption per frame. Even though these tests were performed with minimal interference of other running processes, to minimize the impact of kernel processes on the intended measurements, the results were obtained running the same test several times (e.g., [6][7][8] and then normalising the average results. Also to get more accurate measurements, these started at each frame's first call of the decode method (class TDecTop) and finished after the execution of loop filters, when frames are fully reconstructed.…”

“…For instance, the performance of H.264/AVC decoding was investigated in [5] to identify the main bottlenecks and to improve decoder architectures. Closer to the scope of this paper, a power consumption model for H.264/AVC video c 2018 IEEE decoders with hardware accelerator, was presented in [6]. The authors show that a small number of parameters can be used to estimate the power consumption of hardware accelerators with a maximum prediction error of 10%.…”

Power aware HEVC streaming for mobile

“…Cho et al [8] applied DVFS to the video playback system. Li et al [4] performed power consumption analysis for video decoding based on video characteristics such as spatial and temporal resolution and video quality. MPEG recently started to work on the standardization of green MPEG technologies [9] In this paper we propose a power aware HEVC streaming system based on DASH to improve the quality of experience on mobile platforms.…”