Abstract
The continual development of technologies of creation and diffusion has resulted in an enormous amount of multimedia document. Thus, efficient retrieval and management of these audiovisual document is needed. Therefore and in order to operate efficiently, audio and video browsing deploy automatic processes like speech/music detector systems. In this paper, a new concept for audio characterization based on speech/music discriminator was presented using a blind audio watermarking scheme. In fact, the class of each audio segment is embedded by watermarking. Consequently, browsing or monitoring systems can recognize audio class, i.e. speech or music, at each moment just by extracting the corresponding watermark.