Cochlear implants are prosthetic devices that deliver direct electrical stimulation to auditory nerve fibers as a mechanism to restore normal hearing in deaf individuals. These devices have resulted in a tremendous improvement in the quality of life for individuals who in some cases have been deaf for decades. Moreover, cochlear implants have been shown to provide interpretable auditory information to congenitally deaf children. The underlying mechanisms of cochlear prosthetic-aided hearing are incompletely understood, with the sound quality and speech intelligibility generally being far inferior to that experienced by normal hearing individuals. Our research involves characterization of critical underlying mechanisms that impede speech recognition, and developing strategies to improve the presentation of auditory information to implanted individuals.
An implant consists of an array of electrodes that are inserted surgically into the cochlea. The spatial arrangement of electrodes is used to access the spatial arrangement of the nerve fibers normally present in an unimpaired cochlea. After the implantation surgery, an audiologist fits the speech processor for each individual using methodologies similar to those used to fit eyeglasses or hearing aids. However, unlike eyeglass fitting procedures, no prescriptive fitting procedure exists that is guaranteed to fully correct auditory deficits. Substantial variability in individuals' abilities to understand speech with a cochlear implant exists, and very little is known precisely about the causes of this variability. Factors such as electrode placement, electrode design, nerve survival, speech processing strategy, and complex and/or unexpected current paths that result in electrode interactions may play a role in this variability. These factors are difficult, if not impossible, to either monitor or manipulate in persons with cochlear implants. An additional factor may be the fundamental difference in the way that neurons respond to electrical as opposed to acoustic stimulation . The neural response to electrical stimulation is much more deterministic than the neural response to acoustic stimulation. To complicate these issues, performance most likely involves a complex relationship between the patient and specific device parameters. Our research focuses on experimental measures that can be used to modify the fitting procedure to improve subsequent speech recognition, and on novel information coding techniques that can be integrated with modern cochlear implant speech processing architectures to also improve speech recognition.
Our earliest cochlear implant research involved experimentally evaluating spatially-based electrode interactions and investigating their relationship to speech recognition in cochlear implant subjects. More recently, we have extended this work to consider temporally-based electrode interactions and the relationship between spatial and temporal interactions. Most recently we have focused on novel methods for coding the speech signal and for tuning the implanted devices. We have expanded our research area to encompass two additional research areas in cochlear implants, each of which utilizes a physical model in conjunction with the experimental and/or theoretical signal processing analyses. In the first area, we are using acoustic models of cochlear implants to evaluate various experimental hypotheses and to test the efficacy of proposed remediation strategies. Our research employs acoustic models of cochlear implants, which allow carefully controlled testing to be performed in normal hearing subjects and avoids several confounding effects associated with experimentation with persons with cochlear implants. The results of these experiments are being used to guide our experimental cochlear implant research, and have the potential to provide value to the clinical research community. The second research area involves using statistical signal processing techniques in conjunction with a computational model of the neural response to electrical stimulation to explore alternative signal coding paradigms for cochlear implants. We have demonstrated that this technique can successfully predict pre-existing psychophysical data, can reveal previously unknown trends in psychophysical data, and that alternative signal coding strategies may be used to alter the neural response patterns associated with electrical stimulation. The signal coding techniques explored in this research should provide either more natural neural responses, or a more complete representation of the acoustic signal, which may result in better speech recognition for individuals with cochlear implants.