Simulative Analysis of an Adaptive Control Mechanism for Packetized Voice Across the Internet

Abstract

Proposed new control mechanisms for the transmission of packetized voice across the Internet utilize the presence of silent intervals in conversational speech in order to dynamically adapt the behaviour of the audio application to the network fluctuating traffic conditions so as to minimize the effect of packet loss and varying delays on the quality of audio delivered to the destinations. An accurate model of the on-off characteristics of the conversational speech is thus necessary to analyze the performance of those audio communication systems. In this paper, an eight-state Markov model of voice activity in conversational speech has been used in order to assess the adequacy of an Internet audio mechanism that dynamically sets the playout delay value of packet audio in Internet voice-based connections. Based on this model, several simulation experiments have been carried out that show that a sufficient number of silence periods (of sufficiently long duration) occur in a typical human conversation that permit an adequate application of the proposed audio mechanism. In addition, a number of simulative/experimental trials are reported that show that the proposed Internet audio mechanism strikes a favourable balance between the average playout delay and the packet loss percentage experienced during audio conversations over the Internet.