Auditing the content of audio as transmitted by radio-stations is of great interest for governments, publicists, and for managers of radio-stations among others. Our approach consists of making use of a robust audio-fingerprint for characterization of the monitored audio and a proximity index for fast search of the most similar piece of audio among the collection of audio known to the system (ads mainly). Since the audio signal as broadcasted via Internet suffers little degradation, an inverted index proved to be a great solution while pivot based indexes such as the Burkhard-Keller tree and the Fixed Query array turned out to be of no use for our purpose due to the curse of dimensionality. The implemented system performed really well having a 100% recall and it it fast enough to allow real time monitoring of several radio-stations simultaneously with a single desktop computer.
展开▼
