Effects of Cochlear Hearing Loss on the Benefits of Ideal Binary Masking

Ideal Binary Masking (IdBM) is considered as the primary goal of computational auditory scene analysis. This binary masking criterion provides a time-frequency representation of noisy speech and retains regions where the speech dominates the noise while discarding regions where the noise is dominant. Several studies have shown the benefits of IdBM for normal hearing and hearing-impaired listeners as well as cochlear implant recipients. In this study, we evaluate the effects of simulated moderate and severe hearing loss on the masking release resulting from IdBM. Speech-shaped noise was added to IEEE sentences; the stimuli were processed using a tone-vocoder with 32 bandpass filters. The bandwidths of the filters were adjusted to account for impaired frequency selectivity observed in individuals with moderate and severe hearing loss. Following envelope extraction, the IdBM processing was then applied to the envelopes. The processed stimuli were presented to nineteen normal hearing listeners and their intelligibility scores were measured. Statistical analysis indicated that participants' benefit from IdBM was significantly reduced with impaired frequency selectivity (spectral smearing). Results show that the masking release obtained from IdBM is highly dependent on the listeners’ hearing loss.