|
|
| Fall 2001: CLSP Seminar Series | Tuesday, May 13, 2008 |
Cochlear implants have been shown to be extremely helpful to postlingually deafened listeners. These devices are excellent sensory aids to lipreading, and many patients obtain enough phonetic information from their devices to achieve significant levels of auditory only open-set word recognition. However, our basic understanding of the mechanisms that CI users employ to perceive speech still lags behind the substantial clinical benefit that they receive. In particular, we do not know the exact process that occurs in a listener's brain that allows him to identify a speech sound based on the acoustic input. However, recent progress in our understanding of these mechanisms allows us to propose, for the first time, a clearly defined family of quantitative models of the identification of speech sounds by CI users. The Multidimensional Phoneme Identification (MPI) models aim to predict the identification of speech sounds by individual cochlear implant users, based on their auditory discrimination abilities along relevant perceptual dimensions. The basic structure of the MPI model is a multidimensional extension of the Durlach-Braida model of intensity perception, incorporating an internal noise model and a decision model to account separately for errors due to poor sensitivity and response bias. In the MPI model each speech sound is associated with a multidimensional Gaussian probability distribution whose spread is dependent on the individual listener's auditory discrimination capabilities. Overlap among these distributions determine the pattern of sound confusions that are predicted for the individual listener. The MPI model provides a complete quantitative description of how listeners encode and combine acoustic cues, and how they use this information to determine which sound they heard. Thus, it allows for testing specific hypotheses about phoneme identification in a very stringent fashion. The MPI model has already provided explanations for a variety of experimental findings in speech perception by CI users, such as individual patterns of vowel and consonant confusions. This research is significant because the MPI model may help improve our understanding of speech perception by CI users (and speech perception in general) to qualitatively new levels. In addition, it is hoped that the MPI model may help develop more theoretically-based guidelines for clinical practice, and thus help increase the average levels of speech perception achieved by CI users. After presenting the general theoretical framework, several examples of application of the MPI model will be discussed.Supported by NIDCD (R01-DC03937), NOHR and DRF.
Biographical Information
Dr. Mario Svirsky is on the faculty of the Department of Otolaryngology, School of Medicine and the departments of Biomedical Engineering and Electrical Engineering at Indiana University - Purdue University Indianapolis Dr. Svirsky's areas of research are in speech production, speech perception and language development of pediatric cochlear implant users and speech processing strategies for cochlear implants. His two main research topics at present are: 1) the development of quantitative models of speech perception by cochlear implant users, and 2) the comparison of language development in children with cochlear implants, hearing aids, or normal hearing.
| The Center for Language and Speech Processing The Johns Hopkins University 3400 North Charles Street, Barton Hall Baltimore, MD 21218 | |||||
 | Telephone: (410) 516-4237 | | Fax: (410) 516-5050 | | E-mail: clsp@clsp.jhu.edu |