Methods for Speech Intelligibility Enhancement

Abstract

Modulation domain has been reported to be a better alternative to Frequency domain for speech enhancement, as speech intelligibility is closely linked with the modulation spectrum. This suggests that the modulation spectrum may assist in the demarcation of speech and noise. Motivated by that, this thesis investigates the role of modulation domain towards the speech intelligibility improvements. Acknowledging the fact that the Gaussian assumption for all noise DFT coefficients does not necessarily hold, this thesis suggests the best noise density function which is suitable for both the frequency and modulation domain based speech applications.