Systematic evaluation of perceived spatial quality in surround sound systems

Sammanfattning: The aim of this thesis is to elaborate a method for the evaluation of perceived spatial quality in surround systems with emphasis on how to find and validate relevant attribute scales originating from verbal elicitation of personal constructs. The main question raised in this work was: Could a method utilising elicited personal constructs be used for finding a set of attributes with sufficient common meaning, thus enabling a group of experienced listeners to make significant judgements of spatial quality in surround sound systems? To investigate this, a method comprising a number of steps was employed. Initially, an elicitation experiment employing elements of the Repertory Grid Technique was performed. In the experiment, verbal data in the form of personal constructs were elicited from experienced listeners considering sounds reproduced through loudspeakers. Each listener rated the sounds on his/her personal constructs. The ratings and the constructs were structured by means of multivariate techniques and Verbal Protocol Analysis to find principal structures in the data set. These structures were examined for their meaning and they were subsequently used for deriving attributes on which scales were defined. The scales were verified in a validation experiment where a new group of experienced listeners considered a number of sound stimuli, which were rated on the attribute scales. The ratings were analysed for statistically significant differences between the stimuli. A new experiment was performed using a 5-channel surround sound system, on a new set of stimuli, in which refined attributes from the previous experiment supplemented by attributes from a new elicitation were utilised. The results showed that Personal constructs could be elicited from listeners in the context of reproduced audio A common perceptual pattern existed, expressed as attributes A list of attributes could be successfully derived Attribute scales derived from personal constructs enabled a group of listeners to make statistically significant judgements of spatial quality on reproduced sounds. It is concluded that a method utilising elicited personal constructs can be used for finding a set of attributes with sufficient common meaning, thus enabling a group of experienced listeners to make significant judgements of spatial quality in surround sound systems.