@inproceedings {Zamaninezhad1626_2016,
	year = {2016},
	author = {Zamaninezhad, Ladan and Hohmann, Volker and Büchner, Andreas and Jürgens, Tim and ,},
	title = {Speech Intelligibility Benefits in a Model of Cochlear Implant Listeners with Ipsilateral Residual Acoustic Hearing},
	booktitle = {Assoc. Res. Otolaryng. MidWinter Meeting (ARO)},
	URL = {http://c.ymcdn.com/sites/www.aro.org/resource/resmgr/Abstract_Archives/UPDATED_2016_ARO_Abstract_Bo.pdf},
	abstract = {Introduction
Speech perception in users of conventional cochlear
implants (CI) degrades considerably in the presence of noise.
Due to advances in surgical techniques, for a subgroup of
CI candidates whose apical auditory nerves can still be
stimulated acoustically, there is the possibility to preserve
acoustic hearing even after implantation. Several clinical and
vocoder studies reported on speech intelligibility benefits for
electro-acoustic (EA-) listening condition in comparison with
electric-only or acoustic-only conditions (termed EA-benefit)
in these listeners.
Goals of the study
The goal of this study is to introduce an auditory model of
speech intelligibility that can predict the EA-benefit. The
model is used to assess the effect of different physiological
factors on speech perception of electro-acoustic listeners.
Methods
Two different physiologically inspired auditory models are
used to simulate the auditory nerve spiking pattern in response
to electric and acoustic stimulation. The auditory model of
Fredelake and Hohmann (2012) produces auditory nerve
spikes in response to electric stimulation while the Meddis
(2006) model simulates auditory nerve spikes in response
to acoustic stimulation. Both spiking patterns were further
processed by the central stage of the model of Fredelake
and Hohmann (2012) to obtain internal representations of
the stimuli, which may be present in the central part of the
auditory system.
The back-end of the proposed model consists of a standard
GMM/ HMM speech recognition system. The internal
representations were fed into the classifier and the recognition
rate determined the predicted speech reception thresholds
(SRTs) for sentences in stationary speech-shaped noise.
Predicted SRTs were compared to SRTs measured in 22
patients equipped with Nucleus Hybrid-L device.
Results
The model predicted EA-benefit (SRT-difference between
electric-only and electro-acoustic listening condition) of up
to 3 dB, which agrees with the EA-benefit observed in most
of the patients. Changing the amount of residual acoustic
hearing in the model resulted in an EA-benefit even if the
speech intelligibility with acoustic-only stimulation was close
to chance, which is in line with the results of some of the
patients. Increasing the amount of electric field spatial spread
in the model increased SRTs both in electric-only and electroacoustic
listening condition.
Conclusion
The model of speech intelligibility for electro-acoustic listeners
can predict the EA-benefit that is observed in actual patients.
The model has been used to assess the influence of different
physiological parameters on speech reception thresholds
and EA-benefit.
Funding
This work is supported by DFG cluster of excellence EXC
1077/1 "Hearing4All".}
}