Output details
13 - Electrical and Electronic Engineering, Metallurgy and Materials
University of Birmingham : A - Electronic, Electrical and computer engineering
Incorporating the voicing information into HMM-based automatic speech recognition in noisy environments
All current speech technology research in some way exploits the so-called source-filter model of speech. This paper, in one of the world leading speech science and technology journals, presents a new approach to the automatic estimation of the source characteristics of speech signals and its incorporation into automatic speech recognition. For the first time, the source characteristics of speech can be estimated and modeled for individual frequency regions. This work was one of the outcomes of £115k EPSRC ‘first grant’ funding and helped to enable another £356k EPSRC responsive mode funding.