14th International Congress of Phonetic Sciences (ICPhS-14)
San Francisco, CA, USA
This paper provides a detailed analysis of perceptual data collected from the Nemours Database of Dysarthric Speech. The database includes a collection of 740 short nonsense sentences spoken by 10 male speakers with varying degrees of dysarthria that have been marked at the phoneme level. All of the sentences were in the form "The X is Ying the Z", where X and Z were monosyllabic nouns and Ying was a monosyllabic verb root with an "ing" ending. The complete set of sentences consisted of 74 nouns and 37 verbs each produced twice by each talker. For each talker, a minimum of 5 normal hearing subjects listened to the complete set of 74 sentences 12 times for a minimum of 60 presentations of each production of each word. As subjects listened to each sentence, they chose the three words that they thought the talker was trying to produce from a set of 4 to 6 response alternatives for each word. The response alternatives were chosen such that they differed from the target word on a single phoneme. The target phonemes could occur in word initial, word final, intervocalic, or medial positions. In addition to word position, target phonemes for the perceptual analysis were classified by voicing (voiced or unvoiced), primary articulator (lips/jaw, tongue tip, tongue body, glottis), and manner of articulation (stop, fricative, affricate, approximate, nasal, and null).
Bibliographic reference. Polikoff, James B. / Bunnell, H. Timothy (1999): "The Nemours database of dysarthric speech: a perceptual analysis", In ICPhS-14, 783-786.