Soundscape - Content Based Search

We present a method to search for environmental sounds in large unstructured databases of user-submitted audio (www.freesound.org), using a general sound events taxonomy from ecological acoustics. We use a Support Vector Machines algorithm to rank sound recordings according to the taxonomy, and integrate it in a content-based web search prototype.
This video demonstrates how content-based audio classification, compared to text-only queries, might help sound designers when searching for environmental sounds.

Category:

Science & Technology

Tags:

• freesound
• content-based search
• environmental sounds
• automatic classification