Short time transient events are extremely challenging to represent in the transform domain employed by common transform-based codecs used in applications such audio compression. These short-time events last for a duration that is much shorter than a typical data window, and consequently have power distributed throughout the transform domain. Accurate representation of these events in the transform domain requires higher bitrates than usually available. A common solution is to use window switching, where smaller windows are used for short time transient events, but this has a negative impact on the bitrate as well. In this paper, we show that with certain simplifying assumptions, transient reconstruction can be reduced to a tractable problem that is performed in the frequency domain, so that the transient event can be easily mixed in with the representation of the non-transient events. A closed form frequency domain representation for the phase of a transient event is introduced, and it is shown that this can be done in an iterative way that allows for increasingly complex transient structures back in the time domain.
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.