CONTENT-BASED AUDIO SEARCH : FROM AUDIO FINGERPRINTING TO SEMANTIC AUDIO RETRIEVAL

ISBN-13: 9783639134186 / Angielski / Miękka / 2009 / 224 str.

Pedro Cano

Opis

This book is for audio information retrievalpractitioners.It is about audio content-based search.Specifically, it is on exploring promising paths forbridging the semantic gap that currently preventswide deployment of audio content-based searchengines. Music search sound engines rely on metadata,mostly human generated, to manage collections ofaudio assets. Even though time-consuming anderror-prone, human labeling is a common practice.Audio content-based methods, algorithms thatautomatically extract description from audio files,are generally not mature enough to provide the userfriendly representation that users demand wheninteracting with audio content. This dissertation has twoparts. In a first part we explore the strengths andlimitation of a pure low-level audio descriptiontechnique: audio fingerprinting. In the second part, we hypothesize that one of theproblems that hinders the closing the semantic gap isthe lack of intelligence that encodes common senseknowledge and that such a knowledge base is a primarystep toward bridging the semantic gap. We present a sound effectsretrieval system which leverages both low-level andsemantic technologies.

O autorze

Scientist, Entrepreneur and Music Technologist. PhD in Computer
Science and Communication, Telecommunication Engineer. Author
more than 45 publications as well as 5 patents in the audio and
music computing area. Assistant professor at UPF. Recipient of an
ICREA Junior scholarship. Pedro is a founder of the firm BMAT,
and currently the CTO.