A psychoacoustic basis for spectral sharpening

Spectral sharpening of speech has been proposed as a possible method for improving speech understanding by listeners with hearing loss. Animal neural physiological data and cochlear models differ in their predictions as to the potential success of spectral sharpening for improved speech recognition. Data from early implementations of spectral sharpening are also ambiguous. Our recent investigation tested the theoretical viability of spectral sharpening for the detection and discrimination of spectral peaks in broadband noise. Results suggested that spectral sharpening (decrements) surrounding spectral peaks (increments) made those peaks more easily detected and discriminated than were spectral peaks without sharpening. All participants with moderate hearing loss demonstrated benefit from the spectral sharpening for peak detection and discrimination. Implications for speech processing algorithms will be discussed.