Low Delay Error Concealment for Audio Signals
Stephan Preihs, Fabian-Robert Stöter, Jörn Östermann · 2012
This paper is concerned with error concealment in digital audio signals. Two model based methods for signal extrapolation are developed and compared. Among these is the usage of adaptive Kalman filtering which can be processed with zero algorithmic delay in real-time. We also improved an approach which is based on linear prediction by enhancing it with an adaptive parameter usage. The proposed methods, especially the linear prediction approach, perform excellently on maximum gap lengths of about 10ms. On some monophonic signals the concealment can be extended up to 100ms without a loss of perceived quality. Both approaches have been compared and evaluated by an informal listening test.