Automatic Generation of Musical Instrument Detector by Using Evolutionary Learning Method

Created by W.Langdon from gp-bibliography.bib Revision:1.4192

  author =       "Yoshiyuki Kobayashi",
  title =        "Automatic Generation of Musical Instrument Detector by
                 Using Evolutionary Learning Method",
  booktitle =    "10th International Society for Music Information
                 Retrieval Conference",
  year =         "2009",
  editor =       "Keiji Hirata and George Tzanetakis",
  pages =        "93--98",
  address =      "Kobe, Japan",
  month =        "26-30 " # oct,
  keywords =     "genetic algorithms, genetic programming",
  URL =          "",
  size =         "6 pages",
  abstract =     "This paper presents a novel way of generating
                 information extractors that obtain high-level
                 information from recorded music such as the presence of
                 a certain musical instrument. Our information extractor
                 is comprised of a feature set and a discrimination or
                 regression formula. We introduce a scheme to generate
                 the entire information extractor given only a large
                 amount of labeled dataset. For example, data could be
                 waveform, and label could be the presence of musical
                 instruments in them. We propose a very flexible
                 description of features that allows various kinds of
                 data other than waveform. Our proposal also includes a
                 modified evolutionary learning method to optimize the
                 feature set. We applied our scheme to automatically
                 generate musical instrument detectors for mixed-down
                 music in stereo. The experiment showed that our scheme
                 could find a suitable set of features for the objective
                 and could generate good detectors.",

Genetic Programming entries for Yoshiyuki Kobayashi