Concurrent categorical scaling of sound quality measures (CCSM) - a database of psychoacoustic measures roughness, sharpness, tonality, loudness and annoyance of artificial and real sounds

To connect sound quality with other psychoacoustic measures experimentally and for model evaluation, a benchmark set of simultaneously assessed measures with a large set of signals is highly desirable. Hence, two experiments were conducted, in which roughness, sharpness, tonality, loudness and annoyance were assessed for a total of 129 artificial and real sounds by normal-hearing listeners using a categorical scale. The instrumental loudness was equalized for most of the sounds to limit the influence of loudness. With test-retest correlation of r²>0.76 for all attributes, results indicated the reliability of the method. Correlation analyses and a multilinear regression revealed that, besides loudness, sharpness contributed most to annoyance (r²=0.72), while roughness (r²=0.04) and tonality (r²=0.08) did not contribute significantly. The present data set demonstrates the applicability of a categorical scale for the concurrent assessment of psychoacoustical properties and can serve as a basis for further research that is required to accurately model psychoacoustic measures for a broad range of signals.