Towards lyrics spotting in the SyncGlobal project

With music markets shifting, the use of music in video productions has become increasingly important. Our novel research project "SyncGlobal" addresses this global music licensing opportunity. Our goal is to find the best acoustic or semantic matches to any video sequence from large-scale intercultural music catalogs with minimal human effort. One important aspect is the retrieval of music excerpts given a semantic query, where one requirement is the ability to search for certain words or phrases inside songs, that match the theme of the corresponding video production. Consequently, here we present our approach towards spotting of lyrics in music recordings. It is based on statistical analysis of sub-sequence dynamic time warping for a capella singing voice recordings. We present algorithm details accompanied by illustrative examples for this approach.