Audio Fingerprinting: Nearest Neighbor Search in High Dimensional Binary Spaces
Journal of VLSI Signal Processing Systems
Digital Watermarking and Steganography
Digital Watermarking and Steganography
| Hi-index | 0.00 |
The increasing usage of the Internet for thedistribution of audio content and the increasing numberof audio broadcasting stations require new supportingtools and methods for the observation of occurrencefrequencies and of possible copyright-infringements. Thispaper describes a general approach for the identificationof audio titles and its application on Internet observation.The concept of an AudioDNA is developed, allowing ahighly compressed representation of a sequence ofacoustic events. Audio titles are identified by using asequence matching method which determines similaritiesbetween observed and reference AudioDNA stored in adatabase. This method is implemented in an highlyscaleable architecture allowing the identification ofseveral audio titles in parallel. The first results with thissystem are very promising, only highly distorted audiotitles are not identified correctly.