Responsive image

Hi-precision audio in listening tests - also in the browser?

Benjamin Pedersen
In listening tests, detailed sound control is sometimes mandatory down to each individual digital sample value and guarantee is needed that they are not unintentionally altered. At other times, a lesser degree of control is acceptable, if on the other hand test execution becomes less restricted. Detailed control of sound is often possible only under “laboratory” conditions where hardware and software are under complete control and sound pressure levels can be accurately calibrated. On the other hand, if test persons can do listening tests at home, via an internet browser for example, collecting large amounts of data becomes faster and cheaper (no laboratory facilities required, and more persons can do tests in parallel). Online listening tests made possible by the Web Audio API offers great flexibility in test execution, but compromises in precise stimulus control must be accepted. This paper analyzes such compromises by discussing technological limitations of Web Audio API followed by validation measurements of sound playback in popular internet browsers. The measurements show that at the detailed level there are significant differences in actual performance of different browsers and behavior is not always as expected. Finally, a solution is presented where audio presentation is delegated to an external audio presenter for situations where the limitations of Web Audio API are not acceptable.
            
@inproceedings{2017_EA_28,
  abstract = {In listening tests, detailed sound control is sometimes mandatory down to each individual digital sample value and guarantee is needed that they are not unintentionally altered. At other times, a lesser degree of control is acceptable, if on the other hand test execution becomes less restricted. Detailed control of sound is often possible only under “laboratory” conditions where hardware and software are under complete control and sound pressure levels can be accurately calibrated. On the other hand, if test persons can do listening tests at home, via an internet browser for example, collecting large amounts of data becomes faster and cheaper (no laboratory facilities required, and more persons can do tests in parallel). Online listening tests made possible by the Web Audio API offers great flexibility in test execution, but compromises in precise stimulus control must be accepted. This paper analyzes such compromises by discussing technological limitations of Web Audio API followed by validation measurements of sound playback in popular internet browsers. The measurements show that at the detailed level there are significant differences in actual performance of different browsers and behavior is not always as expected. Finally, a solution is presented where audio presentation is delegated to an external audio presenter for situations where the limitations of Web Audio API are not acceptable.},
  address = {London},
  author = {Pedersen, Benjamin},
  booktitle = {Proceedings of the International Web Audio Conference},
  editor = {Thalmann, Florian and Ewert, Sebastian},
  month = {August},
  pages = {},
  publisher = {Queen Mary University of London},
  series = {WAC '17},
  title = {Hi-precision audio in listening tests - also in the browser?},
  year = {2017},
  ISSN = {2663-5844}
}