Graeme Clark Collection

Permanent URI for this collection

http://hdl.handle.net/11343/375

Filters

1986 1
1
Reset filters

Settings
Statistics
Citations
End date: 1989 × Start date: 1980 × Author: Seligman, Peter × Type: Journal Item ×

Search Results

Now showing 1 - 1 of 1
  • Item
    Thumbnail Image
    Speech recognition performance with a two-formant coding strategy for a multi-channel cochlear prosthesis [Abstract]
    Dowell, R. C. ; Blamey, P. J. ; Seligman, P. M. ; Brown, Alison M. ; Clark, Graeme M. ( 1986)
    Over the last two years, a new speech coding strategy (F0F1F2) has been developed for the Nucleus multi-channel cochlear prosthesis designed to provide information about the first formant, in addition to the second formant and voicing frequency information provided by the “standard” speech processing strategy (F0F2). This strategy uses quasi-simultaneous stimulation of two electrode pairs within the cochlea at the voice pitch rate. The positions of the two sites of stimulation vary independently according to the frequencies of the first and second formants. The amplitude at each site is determined from the first and second formant amplitudes. Seven patients were changed to this strategy and an initial study showed significant improvements in recognition of open set sentence material (from a mean of 30.4% for F0F2 to 62.9% for F0F1F2) and for speech tracking without lipreading (from 11.8 wpm to 30.5 wpm). Phoneme recognition investigations indicated that: 1) vowel identification was improved due to the addition of first formant frequency information in the new strategy, 2) consonant identification was also improved, due to the extra information provided by the independent variation of the amplitude components. These encouraging results led to the use of the F0F1F2 strategy for all new patients from April 1985. Results for recorded speech testing (MAC battery) three months after surgery have been compared for 13 patients who used the F0F1F2 strategy. Significant improvements were observed for the F0F1F2 group on most of the tests. Mean scores for open set testing were as follows: a) spondee recognition: 13.6% for F0F2 and 26.0% for F0F1F2, b) CID sentences: 15.9% for F0F2 and 37.8% for F0F1F2, c) monosyllabic words: 4.9% for F0F2 and 12.4% for F0F1F2, d) phoneme recognition: 23.1% for F0F2 and 33.4% for F0F1F2.