Syllabic level automatic synchronization of music signals and text lyrics

We present a framework to synchronize pop music to corresponding text lyric. We refine line level alignment achievable by existing work to syllabic level by using a dynamic programming process. Our main contribution is using music knowledge to constrain the dynamic programming search. This is done by modeling (1) non-uniform note length distribution and (2) a note length distribution for each section type (for example intro, chorus, and bridge). These reduce alignment error by 6.4% and improve time efficiency by a factor of 2.2. Categories and Subject Descriptors H.5.5 [Information Interfaces and Presentation]: Sound and Music Computing – Methodologies and Techniques; H.3.1 [Information Storage and Retrieval]: Content Analysis and Indexing General Terms Algorithms, Design, Experimentation. Keywords Music structure, voice alignment, dynamic programming, hidden Markov model.

• Type: Conference
• Year: 2006
• Where: MM
• Authors: Denny Iskandar, Ye Wang, Min-Yen Kan, Haizhou Li
• Keywords: Dynamic Programming, Line Level Alignment, MM 2006, Note Length Distribution