Listening tests like dissimilarity tests are traditionally used to study sound perception and build perceptive timbre spaces. As the processing of such tests is tedious when a large number of sounds are to be judged, this work proposes a method to automatically compute the dissimilarity between sounds by combining an auditory representation, the cochleagram, and the concept of time-frequency masks. The cochleagram gives an energy distribution in a timefrequency space which can be considered as a perceptive sound representation. By introducing auditory representations in time-frequency analysis, we wish to propose new tools to characterize timbre. Time-frequency masks that carry information about the time-frequency differences between two signals applied to auditory representations might lead to global timbre descriptors that, as opposed to most traditional timbre descriptors, also are adapted to signals that strongly evolve with time. In this study we applied the proposed method to engine sounds that are complex and rich signals whose perception strongly depends on the dynamic sound variation. In order to evaluate the robustness of our model, we compared the timbre space obtained with perceptive tests to the one generated by our proposed method.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.