Scalable Metadata and Quick Retrieval of Audio Signals

Audio search algorithms have reached a degree of speed and accuracy that allows them to search efficiently within large databases of audio. For speed, algorithms generally depend on precalculated indexing metadata. Unfortunately, the size of the metadata follows the same exponential trend as the audio data itself, and this may lead to an exponential increase in storage cost and search time. The concept of scalable metadata has been introduced to allow metadata to adjust to such trends and alleviate the effects of forseeable increases of data and metadata size. Here, we argue that scalability fits the needs of the hierarchical structures that allow fast search, and illustrate this by adapting a state-of-the-art search algorithm to a scalable indexing structure. Scalability allows search algorithms to adapt to the increase of database size without loss of performance.

• Type: Conference
• Year: 2005
• Where: ISMIR
• Authors: Nancy Bertin, Alain de Cheveigné
• Keywords: Information Retrieval, ISMIR 2005, Metadata, Precalculated Indexing Metadata, Search Algorithms