Time-domain digital time-scale modification algorithms based on the overlap-and-add technique were developed mainly for speech signal applications and may not perform equally well for broad-band audio signals. Broad-band audio signals in general do not have a single 'pitch period' or similar structure to take advantage of, as exists in single-talker speech. Due to the basic assumption of quasi-periodicity in such methods, this means that the best alignment point between two successive frames in the overlapping region may not always be ideal. It results in a distortion in the form of repetitive waveforms, which produce echo and stuttering effects that are audible, especially after time-scale expansion. To overcome this problem, the broad-band signal is first divided into smaller subbands before the overlap-and-add technique is applied. The proposed subband analysis synchronous overlap-and-add (SASOLA) algorithm provides an alternative for time-scale modification of broad-band audio and speech signal applications without the various undesirable artifacts and high computational cost associated with the frequency-domain techniques.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.