Sound samples metadata are usually limited to a source label and several related textual labels. In the context of sound retrieval this makes the retrieval of sounds having no identifiable source (?abstract sounds?) a hard task. We propose a description framework focusing on intrinsic perceptual sound qualities, based on Schaeffer?s research on sound objects, which could be used to represent and retrieve abstract sounds and to refine traditional search by source for non-abstract sounds. We show that some perceptual labels can be automatically extracted with good performance, avoiding the time-consuming manual labelling task, and that the resulting representation is evaluated as useful and usable by a pool of users.
(See document for exact affiliation information.)
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 Join the AES. If you need to check your member status, login to the Member Portal.
