Evaluation of the Perceived Speech Quality for G729D and Opus : With Different Network Scenarios and an Implemented VoIP Application

Abstract: Communication has always been a vital part of our society, and day-to-day communication is increasingly becoming more digital. VoIP (voice over IP) is used for real-time communication, and to be able to send the information over the internet must the speech be compressed to lower the number of bits needed for transmission. Codecs are used to compress the speech, or any other type of data transmitting over a network, which can introduce some noise if lossy compression is used. Depending on the bandwidth, bit rate, and codec used can distortion be minimized which would result in higher perceived speech quality. In the thesis, two codecs, G729D and Opus, were tested and evaluated with two different objective perceive speech quality metrics, POLQA and PESQ. The codecs were also tested with different emulated network scenarios, 2G, 3G, 4G, satellite two-hop, and LAN. Furthermore, Opus was tested with and without VAD (voice activity detection) to see how VAD could affect the perceived speech quality. The different network scenarios did not impact the results of the evaluation, since the main difference between the network scenarios was latency, which POLQA and PESQ do not consider in the evaluation. Opus achieved a higher MOS-LQO (mean opinion score listening quality objective) than G729D. However, when VAD was enabled with Opus for a low bit rate, 8 kbit/s, the MOS-LQO was lower than without VAD.