Lowering mean fundamental frequency to improve speech intelligibility in noise under simulated electric‐acoustic stimulation.

Abstract

We have previously demonstrated that much or all of the benefits of electric‐acoustic stimulation (EAS) can be achieved when the low‐frequency speech is replaced with a tone that is modulated in both frequency and amplitude with F0 and amplitude envelope cues derived from the target speech. One advantage of this approach is that the frequency of the carrier tone can be lowered with little decline in benefit. Lowering mean F0 has the potential to provide EAS benefit to CI users who have very limited residual hearing. One drawback to this approach is that it relies on the efficacy of the pitch extraction algorithm. This is problematic because pitch extractors have trouble in background noise, an environment in which F0 is particularly useful. Here, an alternative way of lowering mean F0 that is unaffected by the presence of noise is examined under simulated EAS conditions. Speech intelligibility was measured using an algorithm based on resampling and compared to performance with the pitch‐based method we have used previously, at frequency shifts of 0, 0.5, and 1 octave. At the 0.5‐octave shift, the resampling‐based approach provided more benefit than the tone. However, at the 1‐octave shift, resampling was less beneficial. [Work supported by NIDCD.]