Abstract for robinson_esca92

Proc. ESCA workshop: Comparing Speech Signal Representations

THE STATE SPACE AND "IDEAL INPUT" REPRESENTATIONS OF RECURRENT NETWORKS

Tony Robinson

April 1992

This paper looks at the data representations used in recurrent networks for two of the supplied sentences for the workshop. One sentence from the database on which the network was trained (timit) is used to illustrate the input, state and output representations for clean speech. Another sentence (clean) is used to illustrate the degradation that results from different recording conditions. Gradient descent in the input space is used on the second sentence so as to make the output better conform to the assumed pronunciation.


(ftp:) robinson_esca92.ps.Z (http:) robinson_esca92.ps.Z
PDF (automatically generated from original PostScript document - may be badly aliased on screen):
  (ftp:) robinson_esca92.pdf | (http:) robinson_esca92.pdf

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.