INTERSPEECH 2013
14thAnnual Conference of the International Speech Communication Association

Lyon, France
August 25-29, 2013

Performance of the MVOCA Silent Speech Interface Across Multiple Speakers

Robin Hofe (1), Jie Bai (2), Lam A. Cheah (2), Stephen R. Ell (3), James M. Gilbert (2), Roger K. Moore (4), Phil D. Green (4)

(1) Independent Researcher, The Netherlands
(2) University of Hull, UK
(3) Castle Hill Hospital, UK
(4) University of Sheffield, UK

This paper investigates the performance of a silent speech interface (SSI) based on permanent-magnetic articulography (PMA) across several speakers. In a previously published study, the SSI was shown to be capable of distinguishing between voiced and unvoiced plosives ([b,p] and [d,t]) in data recorded from a single speaker; a surprising result in a system without access to speech acoustics. The study presented in this paper verifies this finding for three male native English speakers. Furthermore, the speech data used in this investigation was designed to shed light on the capability of the SSI to differentiate between a wider range of phones. A particularly interesting result is that it is possible to distinguish between voiced and unvoiced fricatives ([s,z] and [f,v]) as well as plosives, using PMA data.

Full Paper

Bibliographic reference.  Hofe, Robin / Bai, Jie / Cheah, Lam A. / Ell, Stephen R. / Gilbert, James M. / Moore, Roger K. / Green, Phil D. (2013): "Performance of the MVOCA silent speech interface across multiple speakers", In INTERSPEECH-2013, 1140-1143.