OLT

Optical Logo-Therapy

Short Description

Visualisation of the Articulation with Neural Network Phonetic Maps.

Contact

Athanassios Hatzis
University of Sheffield
Dept. of Computer Science
Speech Group
Regent Court
211 Portobello Street
Sheffield S1 4DP, UK,

email: n.hatzis@dcs.shef.ac.uk

World Wide Web

OLT

Athanassios Hatzis

Description

OLT provides visual feedback to the subject by representing an utterance as a trajectory on a 2D 'phonetic map'. The map is created by Sammon mapping, a technique which defines, for a given set of N-D vectors, a non-linear projection into a 2-D space. The data set for which a Sammon map is created is obtained by Kohonen's Learning Vector Quantisation, LVQ, from a training set of utterances from normal speakers. The path plotted on the Sammon map for a new utterance provides a qualitative comparison against a standard of speech performance. OLT also contains tools to make quantitative comparison and display the results. Finally OLT can be specialised to deal with particular speech training problems.