Efficient Voice Communication inWireless Packet Networks

We present a methodology to optimize the system parameterization for speech and audio transmission in heterogeneous packet networks with wireless access. To this end, we determine the theoretical performance of forward error correction (FEC) schemes on packet level, i.e., the resulting frame loss distribution at the receiver after error correction. The calculation utilizes a statistical model for packet loss distribution which is adaptable for different packet transmission time intervals and packet sizes. The optimal parameterization of the transmission scheme, i.e., the encoding rate, frame length, and amount of redundancy per packet, are then found by optimizing the resulting conversational quality. We utilize the ITU-T E-model which provides a quality prediction in dependence on codec, end-to-end delay, and frame loss distribution. The procedure will be illustrated for realistic transmission scenarios of wireless packet network access technologies.