Automatic Musical Genre Classification of Audio Signals.

摘要: Musical genres are categorical descriptions that used to describe music. They commonly structure the increasing amounts of music available in digital form on Web and important for information retrieval. Genre categorization audio has traditionally been performed manually. A particular musical genre is characterized by statistical properties related instrumentation, rhythmic its members. In this work, algorithms automatic signals described. More specifically, we propose a set features representing texture instrumentation. addition novel strength proposed. The performance those feature sets evaluated training pattern recognition classifiers using real world collections. Based hierarchical classification two graphical user interfaces browsing interacting with large collections have developed.