"Perception & Thresholds of Nonlinear Distortion using Complex Signals" Saw this super long paper posted elsewhere. What do people here make of it?

[Zerimas]

Link to paper.

I was skimming reddit (because apparently I hate myself) and saw this posted on there. I have not read it, but I skimmed the comments of the post (available here). I don't think I am qualified to evaluate the merits of this paper. My "gut" response it is that is a bunch of "scientific" sounding stuff to make the old audiophile claim "measurements don't reveal fidelity", but I could be wrong.

I think probably this quote is the most relevant:

Both the DS and Rnonlin metrics prove to be highly correlated with subjective ratings, and therefore are good metrics in evaluating the perception of nonlinear distortion. While the Rnonlin provides a slightly higher correlation to subjective ratings in the evaluation of artificial distortion, the DS metric also proves to be highly correlated. In contrast, the THD & IMD metrics prove to be highly uncorrelated with sub- jective perception of distortion. -Analysis of the data obtained from the experiment showed that subjects were consistent in their responses and showed a statistical independence between the stimuli presented and subject ratings, indicating that a broad range of distortions were tested. Both the DS and Rnonlin metrics were found to be highly corre- lated with the subjective perception of nonlinear distortion. In contrast, both the THD and IMD metrics were found to be uncorrelated with the subjective perception of nonlinear distortion. With the DS and Rnonlin metrics verified as being correlated with the perception of distortion, the metrics may be used to investigate other perceptual properties of nonlinear distortion. Therefore, the threshold of audibility of nonlinear distortion may be investigated with a metric related to the perception of nonlinear distortion which was never before available with the conventional metrics.

Discussion - Conclusions A verification experiment was conducted to confirm the correlation of the conventional metrics, THD and IMD, and the newer metrics, DS and Rnonlin , with the perception of nonlinear distortion using one music stimuli and four types of nonlinear distortion. It was found that both conventional metrics were not well correlated with subjective perception of distortion with correlations of -0.4845 and -0.4466, respectively. Both of the newly developed metrics, on the other hand, were found to be well correlated with subjective data obtained with overall correlations of -0.9486 and .9547, respectively. From these correlations, it was concluded that the conventional metrics were not well suited to assess the subjective perception of distortion. In contrast, the DS and Rnonlin metrics provide a means for objectively quanti- fying the perceived amount of distortion in a complex music stimulus.

So THD and IMD don't correlate with subjective impressions, but some other scales do? Are THD and IMD even at the levels where it would be audible in first place? Is it even possible to have a "legitimate" subjective impression of these things to begin with when they generally aren't even at audible levels? I don't know. I'd have to read the paper and see exactly what methodology they are using but I am kind of tired at the moment from trying to figure other stuff out. I figured I'd post it here and see what all y'all have to say.

Forgive me if this is a repost.

EDIT: Yes, this paper has been mentioned before. Sorry! Well, it didn't generate too much discussion last time it was mentioned.

 

[sajunky]

Politics...