Speaker-Specific Long-Time Spectrum

M. Sigmund (Germany) and T. Dostal (Czech Republic)


speech processing, long-time spectrum.


In this paper, the average long-time LPC-based spectrum is evaluated as the parametric representation for speaker recognition. Various factors are discussed and experiments were performed as follows: evaluation of spectrum vs. duration of text-independent utterance, comparison of vocal tract spectra obtained from various languages, comparison between intra- and inter-speaker variability in long-time spectrum, and search for a small specially isolated-word vocabulary chosen for Czech language.

Important Links:

Go Back