TODO: HnrAnalysis: Explore better ways to obtain noise floor estimates

ljuvela / GlottDNN

GlottDNN vocoder and tools for training DNN excitation models

Apache License 2.0

32 stars 4 forks source link

TODO: HnrAnalysis: Explore better ways to obtain noise floor estimates #15

Closed mairaksi closed 5 years ago

mairaksi commented 8 years ago

(1) Cepstral processing? (2) Frequency domain "sinusoidal synthesis" (construct harmonic peaks + amplitudes, convolve with analysis window)? (3) Something else? What?

ljuvela commented 8 years ago

We should check that pre-enhancement is applied properly when doing HNR analysis and modification. Is the (differentiated) excitation signal always in the right domain?

mairaksi commented 8 years ago

I believe the pre-emphasized signal (i.e. glottal flow derivative) is always the correct one. Analysis and synthesis are always done within the same domain, since we are not doing integration to glottal flow at any point.