Abstract
This paper presents the extension of an audio watermarking approach to the blind detection of multiple watermarks. The underlying statistical watermarking method working in the fourier domain with different test functions is investigated. The extension of this semiprivate watermarking method to enable a blind detection is presented. To enable the embedding of multiple watermarks the algorithm distributes the different watermarks in individual frequency bands. Several experiments verify the theory and show the different robustness levels of the watermarks. Enhancements and variations of the algorithm are discussed.