Envelope detection based workload prediction for partial decoding scheme

Abstract

In this paper, a novel envelope detection based workload prediction is proposed. The workload estimation is based on the envelope of the difference value between the previous adjacent actual-workloads, and the envelope is obtained by an absolute value and low pass filter method. To further improve the performance of the envelope detection, the negative truncation is introduced by ignoring the negative difference. This workload prediction method is proposed for the partial decoding scheme and works together with the Dynamic Voltage Frequency Scaling (DVFS) to reduce the power consumption of H.264/AVC video decoding. It provides accurate estimation results with only 0.66% deadline missing rate, which is much lower than previous methods. The power consumption results on the evaluation board show that, with the proposed envelope detection based workload prediction and DVFS, the power reduction achieves about 41.65%, the energy reduction is about 10.16%. The energy reduction of the partial decoding together with proposed workload prediction and DVFS is about 61.15% compare to the fully decoding without DVFS.