Parameter generation

Numpy implementation of speech parameter generation algorithm.

build_win_mats(windows, T) Builds a window matrix of a given size for each window in a collection.
mlpg(mean_frames, variance_frames, windows) Maximum Parameter Likelihood Generation (MLPG)
mlpg_grad(mean_frames, variance_frames, …) MLPG gradient computation
unit_variance_mlpg_matrix(windows, T) Compute MLPG matrix assuming input is normalized to have unit-variances.
reshape_means(means, static_dim) Reshape means (T x D) to (T*num_windows x static_dim).