A phonetic contrast function to improve sound processing in hearing aids

Abstract

An objective of a hearing aid (HA) is to improve speech communication for the hearing-impaired listeners. The current best practices for dispensing HAs involve pure tone audiometry (PTA), a functional test that characterizes ability to detect pure tones, but not perception of complex time-frequency patterns in natural speech. We present a phonetic-contrast function (PCF) that takes a test-reference pair from the Minimal Contrast Set (MCS) and generates a time-frequency-energy matrix of contrast regions. The test word is first time-aligned to the reference word using modified dynamic time warping. The frequency axis is composed of 11 bands defined by an audiometric filterbank presented in a companion work. The energy is in a pseudo dB scale in the ±120 dBSPL range. We will demonstrate the operation of PCF for English MCS spoken by a male and a female speaker. In a companion work, we developed a functional test for listening competence based on MCS that include substitutions, deletions, and insertions of acoustic-phonetic features in the stimuli, as in [lock, log, locks]. Future work includes PCF-based investigations with hearing impaired listeners to improve intelligibility of fricatives, stops, etc. by refining the HA parameters in the 11 audiometric frequencies used in sound processing.