Highly efficient and robust audio identification and analytics system to secure royalty payments for song artists

Radio is yet one of the most popular broadcast media in many emerging regions and operates as one of the principle sources of income for licensors and publishers, collectively termed as owners of copyrighted music. Royalty payments are legal obligations towards owners of copyrighted music under intellectual property rights legislations. Independent conjoint monitoring of copyrighted music being broadcast over every radio station on each day is an intricate requisite for upholding the said legislations of a country or region and therefore requires automated techniques that are efficient, scalable and robust to both radio and content noise. We consider the development of an automated radio broadcast audio monitoring system with identification and analytics capabilities to assist collection of royalty payments from radio music licensees for copyrighted music, specifically for songs. First, we pre-process a radio broadcast stream to identify song segments, aka objects via an efficient onset detection mechanism. Next, we perform efficient hashing of identified objects in the frequency domain and compare generated hashes with those in the entries of a pre-compiled copyrighted song database using an efficient hash matching technique. Subsequently, our system stores broadcast data of the identified music objects (e.g. timestamp, name of lyricist, etc.,) in the database for later use from an analytic application.