Aural classification of impulsive sounds

A human listening test is designed to elucidate human perceptions of impulsive sonar sounds. Eleven subjects listen to sound pairs and rate their dissimilarity on a numerical scale. Each subject also verbally describes each sound. The resulting dissimilarity matrices are analyzed to determine measurement repeatability. Data from a subset of subjects is analyzed using Multidimensional Scaling (MDS). The geometric display of the data shows distinct clusters of sounds. Aural inspection of this representation shows that the axes do not appear to represent consistent sound qualities across all clusters. Modeling perception with contrast features as opposed to MDS is presented. Future analysis using this method is outlined and discussed.