Driven by an increasing need for characterizing multimedia material, much research effort has been spent in the field of content-based classification recently. This paper presents a system for automatic identification of audio material from a database of registered works. The system is designed to allow reliable, fast and robust detection of audio material with the resources provided by today's standard computing platforms. Based on low level signal features standardized within the MPEG-7 framework, the underlying audio fingerprint format bears the potential for worldwide interoperability. Particular attention is given to issues of robustness to common signal distortions, providing good performance not only under laboratory conditions, but also in real-world applications. Improvements in discrimination, speed of search and scalability are discussed.
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.