A new technique for audio packet loss concealment

We present a new error concealment technique for audio transmission over heterogeneous packet switched networks based on time-scale modification of correctly received packets. An appropriate time-scale modification algorithm, WSOLA ("Waveform Similarity Overlap-Add"), is used and its parameters are optimized for scaling short audio segments. Particular attention is paid to the additional delay introduced by the new technique. For subjective listening tests, packet loss is simulated at error rates of 20% and 33% and the new technique is compared to previous proposals by category and component judgment of speech quality. The test results show that typical disturbance components of other techniques can be avoided and overall sound quality is higher.