This paper describes methods for automatic extraction of descriptive metadata for audio material and the workflow of archiving. These new algorithms and archiving tools developed at Fraunhofer IMK are to be directly integrated into MediaFabric, a commercially available radio broadcasting framework. Processing steps are based on pattern recognition algorithms and include speech/non-speech detection, speaker change detection and classification, jingle and advertising recognition. The extracted audio structure is described as a hierarchical representation of segment nodes annotated with suitable metadata. An extended retrieval application allows interactive display and navigation of the audio structure. A novel approach to keyword search based on a syllable representation of audio material is used for effective retrieval within the digital radio archive.
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.