Inter-formant Intelligibility Dips for Narrowband Speech

Supported by an EPSRC ROPA

Principal Investigators: Phil Green and Martin Cooke, Research Associate: Stuart Cunningham


Recent speech perception studies have demonstrated high intelligibility of narrow bands of speech (e.g. [2]). While intelligibility is highest in the F2 region, closer inspection reveals intelligibility dips in the region between typical F1 and F2 locations. Experiments conducted in Sheffield have also shown such a dip in intelligibility [1]. Furthermore, a computation model employing missing data techniques has also demonstrated such behaviour.


This study aims to investigate these findings via listening experiments with sentence material. In particular, we plan to investigate how, if at all, the location of these dips is related to the fundamental frequency of the speaker. Also we intended to identify whether dips between F2 and F3, as predicted by the computational model, occur in listeners' results.

Informative references
[1]   Cunningham, S.P. and Cooke, M.P. (1999) The role of evidence and counter-evidence in speech perception, Proc. XIVth International Congress on Phonetic Sciences., San Francisco CA., 215-218.

[2]  Warren, R.M., Riener, K.R., Bashford, J.A, and Brubaler, B.S. (1995) Spectral redundancy: Intelligbility of sentences heard through narrow spectral slits, Percept. Psychophys., 57, 175-182