A PHONETIC TACTILE SPEECH LISTENING SYSTEM
E. M. Ellis and A. J. Robinson
May 1993
Much of the current work on tactile aids concentrates on using the raw acoustic speech signal or close derivatives to stimulate the skin as supplementary information to aid in lip reading. The information rate per skin area associated with many of these schemes is too high to be processed accurately and thus limited in how well they can assist in speech listening. The scheme in this report presents speech information to the skin as phonetic symbols, which exhibits a lower information rate whilst maintaining a high proportion of the original speech information.
As a speech listening system a recurrent error propagation network phoneme recogniser will be the source of phonetic information to drive the tactile display. Phonetic information based on the standard DARPA TIMIT Acoustic Phonetic Continuous Speech Database has been reconstructed via a standard speech synthesiser (MITALK). The results are found to be reasonably intelligible showing that the phonetic format preserves a high percentage of the original speech information. The effects of pitch contours on the resulting speech quality is also investigated and the output from the recogniser is re-synthesised to see how it compares with error-free phonetic information. Suitable hardware has been designed for providing vibratory and pulse stimuli to the skin. Studies and experiments on the sensitivity of the skin show that the data rates associated with the phonetic format are similar to the rates that can be processed through the sense of touch. To complete the scheme, a two-dimensional map of the phonemes of the English language has been formulated for the tactile display. The map is self-organising and exhibits spatial representation according to phoneme similarities.
If you have difficulty viewing files that end '.gz'
,
which are gzip compressed, then you may be able to find
tools to uncompress them at the gzip
web site.
If you have difficulty viewing files that are in PostScript, (ending
'.ps'
or '.ps.gz'
), then you may be able to
find tools to view them at
the gsview
web site.
We have attempted to provide automatically generated PDF copies of documents for which only PostScript versions have previously been available. These are clearly marked in the database - due to the nature of the automatic conversion process, they are likely to be badly aliased when viewed at default resolution on screen by acroread.