Institut

Studium

Forschung


 

The following examples show the results of the use of the whole word models for non-speech signals. Every screenshot exists of the speech signal and the corresponding label file. Each label file consists of 3 traces. The top trace shows the results of the new models, the middle trace shows the former segmentation and the bottom trace shows the hand labelled segmentation.

Positive examples:

Music (INST), applaus (KLATSCHEN)

noise (RAUSCHEN)-breath (ATMEN)-noise (RAUSCHEN)

caught (HUSTEN), noise (RAUSCHEN)

applause (KLATSCHEN)

laughter (LACHEN)-applause (KLATSCHEN)

nosie (RAUSCHEN)-caught (HUSTEN)-noise (RAUSCHEN)

laughter (LACHEN), breath (ATMEN)

Negative examples:
noise (RAUSCHEN)
noise (RAUSCHEN)
complex noise-noise mistake
noise (RAUSCHEN)