Pre-processing

Pre-processing algorithms.

Todo

There’s not common well designed interface yet, may change in near future. https://github.com/r9y9/nnmnkwii/issues/8.

class nnmnkwii.preprocessing.DeltaAppender(windows)

Append delta features.

Given a N x T x D array, features of multiple utterances, transform features into static + delta features for each utterance.

windows

list – A sequence of windows. See nnmnkwii.functions.mlpg() for what window means.

Alignment

class nnmnkwii.preprocessing.alignment.DTWAligner(dist=<function DTWAligner.<lambda>>, radius=1, verbose=0)

Align feature matcies

dist

function – Distance function

radius

int – Radius

verbose

int – Default is 0

class nnmnkwii.preprocessing.alignment.IterativeDTWAligner(n_iter=3, dist=<function IterativeDTWAligner.<lambda>>, radius=1, verbose=0)

Align feature matcies iteratively using GMM-based feature conversion

n_iter

int – Number of iterations.

dist

function – Distance function

radius

int – Radius

verbose

int – Default is 0

F0

interp1d(f0[, kind])

Coutinuous F0 interpolation from discontinuous F0 trajectory