Statistical representation of sound textures in the impaired auditory system

Many challenges exist when it comes to understanding and compensating for hearing impairment. Traditional methods, such as pure tone audiometry and speech intelligibility tests, offer insight into the deficiencies of a hearingimpaired listener, but can only partially reveal the mechanisms that underlie the hearing loss. An alternative approach is to investigate the statistical representation of sounds for hearing-impaired listeners along the auditory pathway. Using models of the auditory periphery and sound synthesis, we aimed to probe hearing impaired perception for sound textures – temporally homogenous sounds such as rain, birds, or fire. It has been suggested that sound texture perception is mediated by time-averaged statistics measured from early auditory representations (McDermott et al., 2013). Changes to early auditory processing, such as broader “peripheral” filters or reduced compression, alter the statistical representation of sound textures. We show that these changes in the statistical representation are reflected in perception, where listeners can discriminate between synthetic textures generated from normal and impaired models of the auditory periphery. Further, a simple
compensation strategy was investigated to recover the perceptual qualities of a synthetic sound texture generated from an impaired model.