A software system for content-based identification of audio recordings is presented. The system transforms its input using a perceptual model of the human auditory system, making its output robust to lossy compression and to other distortions. In order to make use of both the instantaneous pattern of a recording's perceptual features and the information contained in the evolution of these features over time, the system first matches fragments of the input against a database of fragments of known recordings. In a subsequent step, these matches at the fragment level are assembled in order to identify a single recording that matches consistently over time. In a small-scale test the system has matched all queries successfully against a database of 100 000 commercially released recordings.
https://www.aes.org/e-lib/browse.cfm?elib=12998
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.
Learn more about the AES E-Library
Start a discussion about this paper!