In this paper we present a new method for robust identification of highly distorted audio material . The method maybe, e.g., used for queries submitted by holding a cellular phone in front of a loudspeaker . Our approach is basedon our existing audio indexing technology which is suitable for large scale audio data bases. In contrast to recentapproaches, an arbitrary segment of an audio track exceeding a specific length of some 10-20 seconds is sufficient toidentify the corresponding piece of audio. Additionally, the exact position of the fragment within the original signalis determined. In our paper we give an overview on our tests using a data base of approx. 10,000 full-size audiotracks .
展开▼