Iterative Harmonic Speech Enhancement

Several iterative approaches have been proposed for speech enhancement. This work reviews these methods and further presents a novel iterative estimation scheme to jointly estimate the harmonic parameters, i.e., amplitude, frequency and phase of a speech signal observed in noise. The potential of the proposed iterative method in terms of harmonic parameter estimation is analyzed on synthetic harmonic signals. Further, by applying our proposed method on noise-corrupted speech files at different signal-to-noise ratios, we demonstrate its effectiveness in improving instrumentally predicted speech intelligibility and perceived speech quality.