This paper introduces waveform distribution moments as features for audio content analysis. Moments and central moments of distributions are directly calculated from the squared waveform, in order to extract information on the energy development of a signal. The feature trajectories thus obtained promise to be applicable in transient detection, onset detection, and related tasks and are more sensitive to rapid changes than root mean square based methods, as a qualitative analysis reveals. An evaluation of the proposed features is presented in a feature ranking experiment related to transient detection and in an onset detection experiment. In both applications the waveform distribution moments show promising results in comparison to other signal descriptors.
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.