Acoustical analysis on the performance of digital hearing aids

Abstract

The acoustic spectral distances were computed for phonemes, /p/, /t/, and /k/ recorded under a variety of recording conditions, which include: a low (50 dB SPL) and high (80 dB SPL) signal level, a quiet and noisy listening environment, and no processing and processing by three different signal-processing algorithms of a hearing aid. All recordings were obtained using a Zwislocki coupler inside the Knowles Manikin for Acoustic Research. The purpose of this study was to determine whether the acoustic distance between confusable phonemes can be used to evaluate the performance of speech-processing algorithms used in nonlinear, digital hearing aids. The acoustic distance was the rms amplitude difference in the short-term spectra between any two speech segments after they were optimally aligned in time. It was hypothesized that: (1) There would be a significant reduction of acoustic distance among the selected phonemes when speech signals were filtered to simulate a given hearing loss configuration. (2) The use of a hearing aid would significantly alleviate the reduction of acoustic distance among these phonemes. Results, in general, were supportive to these hypotheses. It is, therefore, feasible to assess the information-processing function of a nonlinear, digital hearing aid using an instrument-based approach.