Compensation for clock skew in voice over packet networks by speech interpolation

In voice over packet networks, where the clocks controlling sampling rates at end-points are not always synchronized to each other, small differences in local clock frequencies accumulate over time and result in systematic buffer over/under-runs. This paper describes some signal processing techniques that can be used to compensate for this phenomenon often referred to as "clock skew". The studied algorithms effectively change the number of samples in the receiver play-out buffer dependent on estimated clock skew at the receiver, thereby avoiding systematic over/under-runs of the buffer and allowing shorter de-jittering buffer lengths which, in turn, result in lower end-to-end delay and improved speech quality. The proposed compensation algorithms are based on interpolation techniques.