jAudio is a Digital Signal Processing project built to provide an easy-to-use program for audio feature extraction. Audio feature extraction is extracting properties, such as beat points, statistical summaries, along with many other less obviously useful properties. These properties can then be fed to machine learning toolkits (such as Weka) to automatically extract properties (such as artist or genre) from unknown music. More interesting applications include predicting how much a particular person will enjoy a given piece of music or other hit song predictors.
The application can be used through a GUI, a command-line interface, and with embedding support.
How to cite
jAudio is the end product of a collaboration between Cory McKay and Daniel McEnnis. The core design is from Daniel McEnnis. The core DSP engine is written by Cory McKay. Version 0.3 (published for the ISMIR 2005 conference) integrated the Daniel McEnnis's class project into Cory McKay's original implementation. This version was extended for 0.4 for the ISMIR 2006 conference. Additional extensions related to OMEN were added in 2006.
Development on the system has experienced occasional technical issues, forcing source-only releases after 2008. Development picked back up in 2009 with the release of the jAudio 2.0 specifications document. Additional work on Cory McKay's fork of jAudio (eventually to become jAudio 1.1) for jMir occurs simultaneously. Currently, development has centered on eliminating jAudio library requirements, a 1.1 intermediate release, with the eventual release of a 2.0 system that is expected to eventually merge with the Marsyas project.