Spectro-Temporal Characteristics Affecting the Loudness of Technical Sounds: Data and Model Predictions

Technical sounds are often non-stationary, i. e., their levels vary over time. Thus, temporal as well as spectral characteristics influence the loudness of these sounds. The present study investigated the loudness of technical sounds with a special focus on sounds with pronounced temporal fluctuations. Levels at equal loudness between these sounds and a reference sound were measured for normal-hearing listeners using an adaptive matching procedure. A jet noise was used as the reference sound. In addition to real technical sounds, loudness was measured for stimuli which preserved only the temporal or only the spectral structure of an original technical sound. The results were compared with predictions of published dynamic loudness models and predictions based on the German and American standards for loudness. The accuracy of the models depended strongly on the type of sound. In general, predictions were least accurate for sounds with large temporal fluctuations. The comparison of the loudnesses of the original sounds with those, where only the spectrum or the temporal structures were preserved, revealed that spectral characteristics often dominate loudness. Although none of the tested models could predict the loudness of all employed time-varying, technical signals, the longer time-constants used in one of the models appear to be beneficial.