Improvements in speech processing for the nucleus cochlear implant [Abstract]
Robert Cowan, Richard Charles Dowell, H. D. McDermott, Colette M. McKay, Graeme M. Clark
Abstract
Open-access reader
Robert Cowan, Richard Charles Dowell, H. D. McDermott, Colette M. McKay, Graeme M. Clark
Abstract
Open-access reader
The Nucleus Cochlear implant allows ~ variety of multiple channel speech processing strategies to be developed and trialled. The initial strategy first developed by the University of Melbourne presented the second formant frequency as place and voicing as rate of stimulation. The strategy was subsequently improved by presenting the first format on a place basis as well. Since that time the addition of more spectral information coded as place of stimulation and temporal information presented as variations in amplitude have resulted in better open-set CNC word and sentence scores for electrical stimulation alone. One of the improved strategies selects four pairs of electrodes from the 22 electrode array each glottal pulse to present the first and second formants as well as the output from two high Frequency band pass filters. The other strategy stimulates six of the 22 electrodes representing the six maximal outputs from 16 bandpass filters. The clinical results have shown that both the above strategies results in better open-set speech perception for electrical stimulation in quiet and in noise. 80 % scores have been obtained for open-sets of CNC words and 90 % for open-sets of words in sentences for some of the patients using the latter strategy.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
The Nucleus Cochlear implant allows ~ variety of multiple channel speech processing strategies to be developed and trialled. The initial strategy first developed by the University of Melbourne presented the second formant frequency as place and voicing as rate of stimulation. The strategy was subsequently improved by presenting the first format on a place basis as well. Since that time the addition of more spectral information coded as place of stimulation and temporal information presented as variations in amplitude have resulted in better open-set CNC word and sentence scores for electrical stimulation alone. One of the improved strategies selects four pairs of electrodes from the 22 electrode array each glottal pulse to present the first and second formants as well as the output from two high Frequency band pass filters. The other strategy stimulates six of the 22 electrodes representing the six maximal outputs from 16 bandpass filters. The clinical results have shown that both the above strategies results in better open-set speech perception for electrical stimulation in quiet and in noise. 80 % scores have been obtained for open-sets of CNC words and 90 % for open-sets of words in sentences for some of the patients using the latter strategy.
Key concepts: Cochlear implant, Audiology, Speech recognition, Computer science, Medicine