Comparison of various procedures for the assessment of prominent discrete tones using a large number of sound samples

Abstract

The procedures from the standards ISO 1996-2 and DIN 45681 as well as the two procedures (TNR and PR) from ISO 7779 are compared. In the first section, characteristic features of the procedures (e.g. signal analysis parameters, degree of automation and characteristics of the prominent discrete tone analysis) are compared. In the second section, the procedures are applied to approximately 70 sound samples. Besides natural sounds (e.g. from machines), synthetic sounds (e.g. peak in a trough) are also used. The results are compared against assessments of the sounds by human subjects. The results of the comparisons are discussed, with special attention paid to cases where significant deviations occur. The third section looks at the results of the comparisons under the aspect of requirements for a standard (e.g. robustness, precision). Overall it was determined that none of the procedures investigated provides optimal results for all sound samples.