Hardness is the most commonly searched timbral attribute within freesound.org, a commonly used online sound effect repository. A perceptual model of hardness was developed to enable the automatic generation of metadata to facilitate hardness-based filtering or sorting of search results. A training dataset was collected of 202 stimuli with 32 sound source types, and perceived hardness was assessed by a panel of listeners. A multilinear regression model was developed on six features: maximum bandwidth, attack centroid, midband level, percussive-to-harmonic ratio, onset strength, and log attack time. This model predicted the hardness of the training data with
= 0.76. It predicted hardness within a new dataset with
= 0.57, and predicted the rank order of individual sources perfectly, after accounting for the subjective variance of the ratings. Its performance exceeded that of human listeners.
This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited