SPREI Publicly Available Speech Production Data

One of the objectives of this program is to provide a common articulatory dataset that may be used for the study of many different types of speech production phenomena.

Thee Arctic Articulatory Database (Chilin Shih, Ryan Shosted, Torrey Loucks, Chris Carignan, and Panying Rong) contains Articulograph, audio, and manually-checked time-aligned phone transcriptions for a selection of TIMIT and Arctic sentences. Sentences are selected to guarantee uniphone coverage, and to maximize diphone coverage of English. In order to encourage the speaker to produce interesting types of pronunciation variability, each sentence is recorded three times, with a listener at close, normal, and far from the speaker. Uniphone sentences are recorded nine times each = (3 distances) x (3 speaking styles), where the three speaking styles are accented, clear, and fast.