Physiological correlates of the audibility of masked signals at supra-threshold levels

Hearing entails our ability to perceive individual sounds around us, even in the presence of competing sounds, and it facilitates communication with others. Impaired hearing hampers this ability and can therefore significantly affect quality of life. Hearing cannot be fully restored with current diagnostic measures and solutions, such as hearing aids or cochlea implants. This limitation is likely due to the existing oversimplified models of the complex auditory system underlying the development of diagnostic tools and assistive hearing devices. This thesis aims to address some of the current shortcomings by investigating the auditory processing using three approaches: psychoacoustics, electroencephalography (EEG), and computational modelling.
The sounds we hear are a mixture of multiple frequency components that overlap in time. The auditory system leverages various sound properties to distinguish and group frequency components into relevant target signals and irrelevant background signals. This ability can be experimentally assessed through one’s performance in detecting a target signal in the presence of surrounding noise by varying sound features. The phenomenon behind this is often referred to as masking release, in which the audibility of the masked tone is enhanced when beneficial auditory cues are provided.
By applying psychoacoustic methods, the first part of the study investigates how masking release is induced by three auditory cues present in complex acoustic environments: modulation patterns (comodulation), spatial information (interaural phase difference [IPD]), and the temporal context of signals. These cues facilitate the delineation of frequency components of the target sound from the noise components, thus enhancing target detection performance. Furthermore, the relevance of masking release in speech communication was evaluated. The results show that the IPD cue is beneficial, while comodulation does not significantly improve the audibility.
The second part investigates whether neural responses to the masked tone reflect the audibility of the masked tone. Using EEG, the neural responses evoked by the sound, specifically the late auditory evoked potentials (LAEPs) – are measured in various masking release conditions. The amplitudes of the LAEPs have been correlated with the psychoacoustic measurements of the masked tone’s audibility.
While both psychoacoustic measures and neural responses can be used as audibility measures, these methods cannot provide information concerning the underlying neural mechanisms of sound perception. Thus, in the last part of the study, a spiking neural network (SNN) has been implemented to investigate the neural mechanisms of masking release. The SNN was constructed based on biological neural networks. TheSNN demonstrates potential as a computational tool to explain the neural mechanisms of various psychoacoustic findings.
The results of this thesis provide a comprehensive understanding of auditory processing, which can help improve design strategies for solutions for hearing impairment.