Synthetic speech stimuli spectrally normalized for nonhuman cochlear dimensions

Michael Kiefte, Keith R. Kluender, William S. Rhode
2002 Acoustics Research Letters Online  
Behavioral or neural measures of speech encoding are often taken from animals with auditory systems that differ substantially from those of humans. Absolute distance between spectral peaks of speech sounds along the basilar membrane is typically much greater in humans than in smaller animals. To address this problem, a synthesizer was developed for creating speech scaled for nonhuman cochleae in which spectra are warped to account for differences in cochlear physiology. Absolute cochlear
more » ... ute cochlear distance between spectral peaks (formants) is held constant across species while formant bandwidths are also scaled to span equal numbers of hair cells as for humans. This was accomplished via a flexible reimplementation of the KLATT80 speech synthesizer in MATLAB.
doi:10.1121/1.1445202 fatcat:6rrgaa2vtbb3bikrzbs6mjwkga