next up previous contents index
Next: Sampling rare events in Up: Assessment methodologies and experimental Previous: Experimental design

 

Assessing speaker verification and recognition systems

   

Speaker verification needs to make use of many of the statistical and experimental procedures outlined in previous sections. The principal procedures that are required for assessing speaker verification and recognition systems are those applied to speech recognition processes in the preceding section. The dependent variable is going to be whether the speaker is verified as a legitimate user of the system or recognised as a particular individual, respectively. The outcome is, then, a discrete variable. Discrete variables are somewhat simpler measures than those considered in connection with acoustic measures of speech recognition (note that it is not claimed that the problems involved are necessarily simpler). The particular types of statistical tests called for will be non-parametric (see Section 9.2.6).

As mentioned before, assessing speech recognisers  is integrally related to speech corpora; it is for this reason that an example was worked through from scratch addressing both setting up a corpus and assessing recognition. Here only the additional considerations that need dealing with are described; those to do with sampling speakers for recognition and verification and those to do with assessment.






EAGLES SWLG SoftEdition, May 1997. Get the book...