Optimizing cochlear implants for better perception of speech on speech

Cochlear Implant (CI) devices are electronic prostheses that are implanted in
deaf and hard-of-hearing patients to help restore their auditory sensation.
These implants transduce incoming acoustic signals into pulse sequences that
can be used to stimulate an electrode array implanted within the cochlea. In
spite of the high success rate of CI devices in restoring hearing to profoundly
deaf patients, there exist major limitations to the current clinical
state-of-the-art devices. Despite having high performance on speech-in-quiet
tasks, CI patients exhibit major difficulties in situations in which the target
speech is masked by one or more competing talkers. To segregate competing
talkers in such a situation, normal hearing (NH) listeners make use of voice
differences that are principally supported by two vocal dimensions: the glottal
pulse rate (GPR) and the vocal tract length (VTL). GPR defines the fundamental
frequency (F0) or voice pitch, while VTL is related to the size of the speaker.
It has been shown in a previous study, however, that CI users do not make use
of these two dimensions to identify the gender of a speaker (Fuller et al.,
2015). In contrast to NH listeners, CI patients can only utilize F0 cues and
not VTL to make judgments about the talker*s gender. Moreover, in the SpIdCI
protocol (METc 2012.392), we identified that spectral distortions induced by
the implant (for example as a result of frequency quantization and spread of
excitation across electrodes) are more likely to affect VTL cues than voice
pitch (Gaudrain and Ba*kent, 2015). In the present study we will investigate,
for the first time, how multiple parameters of the CI influence the perception
of these vocal dimensions. Better understanding of how these parameters affect
VTL perception in CI users will allow a more optimal fitting of these devices
for better speech intelligibility in cocktail-party settings.

The main goal of the study is to optimize the perception of vocal
characteristics through a cochlear implant. It will span a number of aspects:
psychophysical characterisation, its relation to speaker discrimination, its
relation to single voice intelligibility, its relation to competing voices
segregation, and its relation to voice emotion recognition. The study focuses
primarily on vocal-tract length (VTL), however F0 will also be manipulated.

The study is an observational study, with both within-subject and across-group
comparisons. In all experiments of the study, a measurement variable is
collected as a function of VTL (and F0), manipulated parametrically. The
subjects are presented with auditory stimuli and given a behavioural task. They
then have to make a judgement and either have to provide an answer on a
computer screen, or they are requested to provide a verbal response. Thus, the
measurement variables will be, in different experiments: percent-correct
discrimination and percent-correct identification. For the tasks in which the
subjects will be requested to respond verbally, their responses are to be
recorded for offline processing. This will only be acquired after they have
given a written informed consent. For CI subjects, they will be required to
perform the tasks while using research speech processors, whose parameters have
been modified according to the study objective of this project. Their own
processors will not be changed in any way. The proposed project consists of 6
experiments spreading over 2 years. Different participants will be recruited
for each experiment, which consists of maximum 2 sessions of 2 hours (NH) or 3
hours (CI). This duration may be extended upon the request of the participants,
or in the case of recording verbal responses. For practical reasons, at the end
of the experiment, they will be invited to participate in other experiments of
the project. For experiment 6, CI participants will be asked to wear a research
speech processor for an extended time period (a few weeks to a few months),
after which they would come for a psychophysical test. For this experiment, 3
sessions will be held: in the first session, their baseline performance while
using their own clinical processors will be evaluated. Additionally, the
research processors will be fitted with the subjects* clinical maps. After the
acclimatization period of a few weeks, they will come for the second and third
sessions that contain the actual psychophysical test.

There is no known risk, nor benefit associated with participation. Given the
absence of risk for the participants* health, the METc of the UMCG has granted
exemption of compulsory insurance for this study. The test session will last
for a maximum of 3 hours per listener, including breaks. Session duration can
easily be accommodated on the request of the participant as needed. The sound
level will always be adjusted to a comfortable listening level for the
participant. Only for Experiment 5 will the subjects be asked to come for 3
sessions, and will be asked to document the duration of use of the research
processor over the course of several weeks.