Speech Modelling and Enhancement

We are interested in the mathematical modelling of speech, primarily in the context of source-filter modelling of speech. These models can be applied, for example, to speech enhancement and glottal inverse filtering.

Specific topics Inverse Filtering Methodology Glottal inverse filtering of the voice, that is, cancellation of the acoustic effect of the vocal tract from a speech signal to obtain information of air flow through the glottis. Speech Enhancement The artificial improvement of the pleasantness and intelligibility of speech signals, especially bandwidth expansion. Speech Processing Mathematics Mathematical analysis of speech processing methods including stability analysis of all-pole models, Line Spectrum Pair polynomials, etc.

More information:

• Speech Analysis team