Standard cochlea implant (CI) speech coding strategies transmit formant information only via the place of the stimulated electrode. In acoustic hearing, however, formant frequencies are additionally coded via the temporal rate of auditory nerve firing. This study presents a novel CI coding strategy (“Formant Locking (FL)-strategy”) that varies stimulation rates in relation to extracted fundamental and formant frequencies. Simulated auditory nerve activity resulting from stimulation with the FL-strategy shows that the FL-strategy triggers spike rates that are related to the formant frequencies similar as in normal hearing, and greatly different than in a standard CI strategy. Vowel recognition in seven CI users via direct stimulation of their electrode array shows that the FL-strategy results in significantly increased scores of the vowels /u/ and /i/ compared to a standard CI strategy. However, at the same time, a decrease in scores for /o/ and /e/ occurred. A microscopic speech intelligibility model involving an automatic speech recognizer reveals good agreement between modeled and predicted confusion matrices for the FL-strategy. This suggests that microscopic models can be used to test CI strategies in the development phase, and gives indications which cues might be used by the listeners for speech recognition.
Cite as: Eichenauer, A., Dietz, M., Meyer, B.T., Jürgens, T. (2016) Introducing Temporal Rate Coding for Speech in Cochlear Implants: A Microscopic Evaluation in Humans and Models. Proc. Interspeech 2016, 635-639, doi: 10.21437/Interspeech.2016-267
@inproceedings{eichenauer16_interspeech,
author={Anja Eichenauer and Mathias Dietz and Bernd T. Meyer and Tim Jürgens},
title={{Introducing Temporal Rate Coding for Speech in Cochlear Implants: A Microscopic Evaluation in Humans and Models}},
year=2016,
booktitle={Proc. Interspeech 2016},
pages={635--639},
doi={10.21437/Interspeech.2016-267}
}