A novel signal processing approach to auditory phantom perception

Abstract

The ear and brain interact in an orchestrated manner to create sensations of phantom tones that are audible to listeners despite lacking physical presence in original sounds. The relative contribution of peripheral sensory cell activity and cortical mechanisms to phantom hearing remains elusive. The current study addressed the question of whether non-linear components of a complex signal exist that are not captured by the linear combination of cosines in a series. To this end, we investigated the source and spectro-temporal dynamics of non-linear components within two-tone complexes related to phantom acoustic perception. The empirical mode decomposition, a method for non-linear and non-stationary processes, was applied to extract the extra-aural existence of an oscillatory component within the original signal associated with the phantom sound. This travelling wave (phantom) has never before been observed in the sound’s linear spectrum. We showed that the wave travels at a velocity that accurately maps onto the perceived phantom tone frequency. Phase coherence of oscillatory mode dynamics predicted discrimination sensitivity to phantom sounds by listeners. Perceived incidences of phantom tones correlated with magnitude of the Hilbert power spectra of the extra-aural component. Findings suggest a possible origin of phantom sounds that exists within the original signal, with potential implications for current models of non-linear cochlear mechanics and cortical dynamics in generating phantom percepts.