globalphone

Multi-layer Perceptrons (MLPs) trained with the GlobalPhone corpus

This archive contains MLPs, and some of the tools to use them, which will provide in their output layers the posterior probabilities for the following labels:

Phonemes

Place of articulation

Manner of articulationNasality

Voicing

Vowel

(vowel) Frontness

(vowel) Height

(lip) RoundingStress

The inputs to the nets are drawn from PLPs - details are given in Section 2.2.1 of https://era.ed.ac.uk/handle/1842/5773. The exact set of values possible for each of the variables above varies with the language used to train the net. The language(s) used were:Mandarin Chinese

German

(Brazilian) Portuguese

Russian

Spanish

Swedish

German, Portuguese and Spanish

Portuguese, Spanish and Swedish