Audio-object segmentation tools

• Silence detection

- Silence audio frames have low energy level with respect to voice

segments and can be discarded by thresholding.

- The average magnitude and zero-crossing rate can be exploited

for end point detection.

• Voiced-unvoiced discrimination

Unvoiced sounds exhibit significant high frequency content in contrast to voiced ones.

Previous slide

Next slide

Back to first slide

View graphic version