Estimation of vowel recognition with cochlear implant simulations

Area of Science:

• Audiology
• Speech Processing
• Biomedical Engineering

Background:

• Cochlear implant (CI) device optimization requires understanding parameter effects.
• Individual patient variability necessitates pre-evaluation parameter estimation.

Purpose of the Study:

• To estimate the effect of CI speech processor parameters on acoustic vowel space.
• To compare acoustic vowel space with normal-hearing subjects' vowel recognition performance.

Main Methods:

• Mel-frequency cepstrum coefficients (MFCCs) estimated acoustic vowel space for CI-simulated stimuli.
• Five CI parameters (spectral resolution, smearing, warping, shifting, amplitude distortion) were simulated.
• Acoustic space was correlated with vowel recognition by normal-hearing listeners.

Main Results:

• Acoustic vowel space highly correlated with vowel recognition for spectral resolution and smearing.
• Correlation was not significant for spectral warping, shifting, or amplitude distortion.
• Vowel recognition was affected by warping and shifting, despite minimal acoustic space changes.

Conclusions:

• CI device may preserve phonetic distinctions under spectral warping and shifting.
• Acoustic analysis is valuable for some CI parameters but not all.
• Auditory training may enhance CI patient perception of speech cues affected by warping and shifting.