Dynamic VoIP codec selection on smartphones

Voice over IP (VoIP) applications can choose a plethora of different speech codecs, which differ in bandwidth, listening speech quality, and resilience to quality degradation under packet loss. However, VoIP Codecs also exhibit differences in facets such as computational complexity or traffic generated that impact on the energy consumption of smartphones due to the use of processor. This work deals with the study of energy consumption differences among VoIP codecs. We compare the execution time required to encode/decode reference conversations. Our results show that computational complexity has a significant impact on battery consumption (a factor of up to 10 was found between different codecs). Based on our results, we provide a ranking of energy efficiency. We also propose a simple algorithm for codec dynamic selection considering the factors of quality, energy and bandwidth. Our algorithm reacts to network conditions choosing the codec that provides less battery consumption constrained to user-defined targets for minimum quality and maximum codec bitrate.