Playout Buffers for VoIP

Project Type: Master/Diploma Thesis
Student: Sakir Emine Zerrin
Mentor: Gernot Kubin


 Voice over IP (VoIP) is a popular and growing technology. Conversational voice quality is an important metric in VoIP applications, which is affected by delay, jitter, and packet loss. Playout buffers can be used in a VoIP system to compensate for network jitter and to resynchronize the received packets. Playout buffers do not play the packets back as soon as they are received but wait for a certain time so that they can be played back in a continuous way. This waiting adds to the overall delay. On the other hand, packets that are received after their scheduled playout time are lost and increase the total loss rate. Because of these reasons, the buffer size is a trade-off between delay and loss. As we want to maximize the conversational speech quality, the best is to make the trade-off between delay and loss based on a quality model. In this work we propose a playout buffer algorithm which is based on maximization of conversational voice quality, aiming to minimize the computational complexity. For this purpose, we first model the network delay using a Pareto distribution and show that this choice is the best compromise between achieving a good fit to the network delay characteristics and yielding a low arithmetical complexity. We use the E-Model as the quality model and simplify its delay impairment function. Our goal is to find the playout delay which minimizes the sum of the simplified delay impairment factor and the packet-loss dependent equipment impairment factor of the E-model. Based on the modified version of the E-model and the Pareto distribution model, we propose a playout buffer algorithm which finds the optimum playout delay using a closed-form solution. The simulation results show that our proposed algorithm is competitive with existing state-of-the-art algorithms with the advantage of a reduced complexity for a quality-based algorithm.