Distortion Perception Results

It would not be professional to post the actual copyrighted papers that we presented to the AES in October, but I don't think that it is unprofessional to publish some of the results from those works. In the tests that we performed, 40 subjects were asked to rate, subjectively, the sound quality of the presentations on a scale of 0-5 where 0-imperceptable, 1-just perceptible, 2-perceptible but not annoying, 3-annoying, 4 very annoying, 5-intolerable.

This figure shows a plot of the average perception of the 21 distorted sound files versus the THD numbers for the same distortion characteristic. The correlation is about -.3, basically there is no correlation. Quite interestingly, there seems to be a fairly high correlation between the inverse of THD and the perception.

This figure shows a plot of the average perception of the 21 distorted sound files versus the IMD numbers for the same distortion characteristic. The correlation is also about -.3, basically there is again no correlation.

This figure shows the correlation for the 18 distorted sound files where the GedLee Metric Gm was less than 10.0. The correlation here is .94 (p<.001) which is outstanding for subjective data. Note that for G_m values less than 1.0, the distortion is basically inaudible. No such threshold number can be found for THD or IMD.

The actual papers can be obtained at www.aes.org

Last modified May 20, 2008 by Earl Geddes