Content-based indexing and retrieval-by-example in audio

This paper presents a system for audio content-based indexing and retrieval-by-example. An unsupervised method to automatically index the audio content is proposed. To realize a retrieval system where the on-line query can be performed efficiently, a new parametric distance metric is used to allow faster identification of the audio segments that are similar to the given query example. Preliminary experimental results show that the proposed new approach can serve as an effective means in content based indexing and retrieval in audio streams