Time scale modification of digital audio signals and its applications

The research work explored the time-scale modification (i.e. contraction arid/or expansion) of digital audio signals and its many applications. The length of digital audio signals is changed in time without changing of their timber or pitch. Tailored to specific applications, we developed several different time-scale modification algorithms, explored the use of them and demonstrated the performance.; One set of the applications is audio/video editing/skimming/playback. Based on our novel modified Waveform Similarity Overlap-and-add algorithm, we could change the length of high fidelity digital audio signals in time without affecting their quality. We also addressed the problem of stereo channel synchronization to avoid sound localization problem encountered when two channels are treated separately.; In packet-based audio transmission systems such as voice over IP, the transmitted audio streams may experience packet loss and delay variation. We developed a modified packet, based SOLA algorithm at the packet level and applied it to adaptive delay concealment for speech streaming applications. We performed time-scale modification based on the estimation of packet delay. The time-scale modification factor is estimated for each packet depending on the delay constraint, delay statistics and the number of late arrival packets. Then, this estimation is bounded by certain upper-bound and lower-bound parameters that can be either fixed throughout the waveform, or calculated based on speech contents. It is shown that the proposed adaptive playout mechanism could be adaptive to the observed delay jitter and/or the loss probability to a certain degree.; With the use of forward error correction and packet-based SOLA, we proposed an adaptive playout algorithm that could lower the overall loss rate of Internet streaming audio without sacrificing the average end-to-end delay. Moreover, we investigated the impact of the packet loss and different error concealment schemes on audio quality. By understanding these impacts, our adaptive algorithm could achieve high perceptual quality with less artifacts caused by the loss, the delay and the stretching algorithm.; We also investigated and developed the objective audio quality measurement scheme that works well for time-scale modification. Thus, it is possible to have an objective measure of the quality of time-scale modified signals.