Responsive image

Using Empirical Analysis of Music Corpora to Optimize Web Audio Playback

Tom Collins, Christian Coulon
Due to feasibility issues and musical preferences, Web audio applications have tended to emphasize the use of synthesized instruments and short samples (e.g., drums) over large banks of longer files that sample other acoustic instruments such as a violin or piano. As the sounds generated by sampled acoustic instruments are quite realistic, they are likely to be of interest to many users of Web audio applications. Using the Tone.js Web Audio framework, this paper describes an initial investigation into load times when rendering music with such sampled instruments. A method is proposed for reducing load times, and hence optimizing Web Audio playback, based on empirical analysis of the note durations used across different music corpora. Experimental results for 400 randomly selected short music excerpts indicate that the proposed method does lead to significant load time reductions, from 3.87 s to 1.72 s. Researchers interested in replicating the results of these experiments or downloading and exploring our playback solution are pointed to http://tomcollinsresearch.net/research/wac/2016/
            
@inproceedings{2016_45,
  abstract = {Due to feasibility issues and musical preferences, Web audio applications have tended to emphasize the use of synthesized instruments and short samples (e.g., drums) over large banks of longer files that sample other acoustic instruments such as a violin or piano. As the sounds generated by sampled acoustic instruments are quite realistic, they are likely to be of interest to many users of Web audio applications. Using the Tone.js Web Audio framework, this paper describes an initial investigation into load times when rendering music with such sampled instruments. A method is proposed for reducing load times, and hence optimizing Web Audio playback, based on empirical analysis of the note durations used across different music corpora. Experimental results for 400 randomly selected short music excerpts indicate that the proposed method does lead to significant load time reductions, from 3.87 s to 1.72 s. Researchers interested in replicating the results of these experiments or downloading and exploring our playback solution are pointed to http://tomcollinsresearch.net/research/wac/2016/},
  address = {Atlanta, Georgia},
  author = {Collins, Tom and Coulon, Christian},
  booktitle = {Proceedings of the International Web Audio Conference},
  editor = {Freeman, Jason and Lerch, Alexander and Paradis, Matthew},
  month = {April},
  pages = {},
  publisher = {Georgia Tech},
  series = {WAC '16},
  title = {Using Empirical Analysis of Music Corpora to Optimize Web Audio Playback},
  year = {2016},
  ISSN = {2663-5844}
}