Development and Validation of a Method for Predicting the Perceived Naturalness of Sounds Subjected to Spectral Distortion

In a previous study it was explored how the perceived naturalness of music and speech signals was affected by various forms of linear filtering. In the present paper a model is introduced to account for the results. The model is based on the assumption that changes in perceived naturalness produced by linear filtering can be characterized in terms of the changes in the excitation pattern produced by the filtering. The model takes into account both the magnitude of the changes in the excitation pattern and the rapidity with which the excitation pattern changes as a function of frequency. It also includes frequency-weighting functions to take into account the fact that naturalness is affected little by changes in amplitude response at very low and very high frequencies. The model accounts very well for the data presented in the earlier study. Two validation experiments were conducted in which naturalness ratings were obtained for speech and music stimuli passed through new sets of linear filters, including filters based on the measured frequency responses of real transducers. The model predicted the results of these experiments well.