Using Autoregressive Models for Real-Time Packet Loss Concealment in Networked Music Performance Applications

In Networked Music Performances (NMP), concealing the effects of lost/late packets on the quality of the playback audio stream is of pivotal importance to mitigate the impact of the resulting audio artifacts. Traditional packet loss concealment techniques implemented in standard audio codecs can be leveraged only at the price of an increased mouth-to-ear latency, which may easily exceed the strict delay requirements of NMP interactions. This paper investigates the adoption of a low-complexity prediction technique based on autoregressive models to fill audio gaps caused by missing packets. Numerical results show that the proposed approach outperforms packet loss concealment methods normally implemented in NMP systems, typically based on filling audio gaps with silence or repetition of the last received audio segment.