Taxonomy of Musical Genres

Many researchers have been conducted to retrieve pertinent parameters and adequate models for automatic music genre classification. It plays a significant role in multimedia applications. In principle, the categorization of music is mostly done by people expert in the field. These are based on several attributes music (timbre, melody, etc.). Despite great efforts employed, the results are very subjective and not very satisfactory. In this work, an ergodic hidden model fully connected is used as one model for 65 musical pieces. Standard Real World Computing (RWC) is used as Database. After training, relative frequency of states transition (histogram) is proposed as a pattern to characterized musical genre. Also, a taxonomy based histogram is presented and compared to manual taxonomy of the RWC.