Spatial Error Concealment in Ad-hoc Audio Conferencing Systems

In this work we consider an ad-hoc audio conferensing system based on VoIP services in which the participants connect to the conference using mobile communication devices with wireless connectivity. To overcome possible quality problems in the wireless link, we propose improvements to the existing conferencing systems. Some networking modifications are suggested to increase the channel capacity and robustness from the conference server to multiple clients. On the other hand, for the improvement of the uplink quality, we suggest a new spatial error concealment method, where a backup device captures and sends the audio signals to the server together with the primary device. In the server the lost frames from the primary channel are estimated based on the backup signal. Several methods for estimating the primary signal based on the backup signal are studied. The results of the methods are evaluated by a psychoacoustic error metric based on Zwicker’s loudness model. An informal subjective test is also performed to compare the results of these methods in order to chose one for implementing on the real- time conferencing setup. Both objective and subjective tests show consistent results and confirm that usage of spatial error concealment improves significantly the audio quality in the primary signal.