Re: spatial separaton (mup1dm )

Subject: Re: spatial separaton
From:    mup1dm  <mup1dm(at)SURREY.AC.UK>
Date:    Fri, 2 Feb 2001 12:39:38 +0000

The group in which I am working is researching a particular artificial environment, that created (or not) by home "surround-sound" entertainment systems. When we set out to make a recording, whether it is a background for a talking-head news report or a music performance in a concert hall, we want to know what choices will improve the "quality" and what choices will degrade it. Our approach so far has been to follow the stimulus-response paradigm, but we have been using several methods to try to get reponses that are meaningful. We have watched the discomfort of industry colleagues who developed headphone spatialisation schemes that performed very well when experimenter asked about localisation, but suddenly performed badly when people started asking about "realism". If we ask subjects to scale this elusive "realism" or "goodness" and also scale the other perceptually relavent attributes of the differences between stimuli, then we hope that we can eventually come up with some optimum points on attribute scales and then at least have somethng to aim at when developing recording and mixing techniques. The trick in this approach is to identify the perceptually relavent attributes for a given population of stimuli. I am in the midst of trying to interpret the spaces spit out by multidimensional scaling, and a colleague is doing well using Rasch repertory-grid-technique. Of course the stimuli have facets of programme type, recording environment, and many facets of recording technique. There are also facets of the playback environment that must be considered, such as possible absence of the centre-front loudspeaker and large loudspeaker frequency-response differences between front/rear/centre-front. The set of most perceptually relavent attributes may differ between a Bach tocatta and a Brandenburg concerto, and may differ between the Brandenburg concerto conducted by Pinnock and conducted by Rattle. The results of tests will never necessarily be generalisable to any programme with a different facet profile. Douglas McKinnie Institute of Sound Recording University of Surrey

This message came from the mail archive
maintained by:
DAn Ellis <>
Electrical Engineering Dept., Columbia University