Compensating for voice over IP network packet loss by insertion of interpolated sub packets

摘要

A receiving endpoint of a packet switched IP network receives audio packets containing a plurality of audio samples which are stored in a jitter buffer where they are treated as a series of sub packets each comprising a smaller number of samples (See Fig. 3). Upon detection of a packet loss the system inserts interpolated sub packets by means of weighted averaging of the samples between the preceding and succeeding sub packets (See Fig.4). The original and inserted sub packets are then played out of the buffer. The size of the sub packets and hence the portion of the buffer that is played out at each time is set to minimise the negative effects on voice quality. The rate of insertion can be controlled to limit frequency shift and stuttering distortions in the output audio signal. The method allows packets to be expanded to compensate for lost packets and results in a relatively low processing burden.