SPIRE VCV: An Acoustic-Articulatory Corpus with Three Different Speaking Rates

We present a corpus, named SPIRE VCV¹¹https://spire.ee.iisc.ac.in/spire/spire-vcv.php which consists of simultaneous acoustic and articulatory data for symmetric vowel-consonant-vowel (VCV) sequences. The recordings were collected from ten non-native English speakers (five female and five male), where articulatory data was recorded using an Electromagnetic articulograph at three different speaking rates, namely slow, normal and fast. The VCV segments comprise five vowels, namely, /a/, /e/, /i/, /o/, /u/ and seventeen consonant sounds, namely /b/, /ch/, /d/, /f/, /g/, /jh/, /k/, /l/, /m/, /ng/, /n/, /p/, /r/, /s/, /t/, /v/ & /z/. The articulatory movements were recorded from six speech articulators-upper lip, lower lip, jaw, tongue tip, tongue body, & tongue dorsum in a 3D coordinate system. The manually annotated boundaries for vowels and consonants are provided with the database. We have also presented a basic articulatory analysis of the consonants contained in the corpus, which describes how articulatory movement changes with the rate. The SPIRE VCV corpus is suited for various acoustic-articulatory studies, particularly for speaking rate specific investigations.