Content Extraction Based on MPEG-7 Audio Representation

摘要

Growth of amount of multimedia data that are stored and used nowadays requires tools for intelligent searching and management. MPEG-7 standard, which is being currently developed, provides unified representation and set of tools for multimedia content description. However, only low-level features of sound signal in MPEG-7 can be extracted automatically, whereas high-level information like sound recognition, musical instrument timbre, etc. requires manual insertion or further data analysis. In this paper, low-level sound descriptors from MPEG-7 standard are a basis for extraction of high-level information. Obtained descriptors can supplement audio data representation and facilitate content-based search of sound and multimedia databases.