Phonetic analysis vs. dirty signals: Fixing the paradox

Abstract:

Noisy speech signals and phonetic data analysis: two phrases that rarely occur in the same context. This paper shows that noise cancellation and noise suppression methods can clean noisy speech signals to such an extent that they can be reliably segmented into phones semi-automatically using appropriate tools such as the Munich Automatic Segmentation Tool. This finding brings noisy signals and phonetic data analysis closer together.


Year: 2013
In session: Werkzeuge und Messverfahren
Pages: 136 to 143