The objective of this research is to develop new sequences that may be more effective for enabling the user of a cochlear implant to recognize sounds than the sequences generated by the present signal processing strategies. These sequences will be derived from the action potential responses found in the VIIIth nerve fibers of a normal system. The new codes that will be obtained in this research will be the average patterns found in an ensemble of fibers and the predominant patterns found in these action potential patterns. The predominant patterns will be obtained through the use of principal component signal processing techniques combined with varimax rotation methods. In order to perform these calculations, the action potential will be transformed into an equally sampled continuous waveforms by low-pass filtering the waveform. The consistency of the average and the predominant waveforms for different subsets of fibers within the same animals and for fibers from different animals will be determined. The linearity of these transformations and the effect on the response of noise added to the stimulus will also be investigated. This research, performed with a model system using very simple sounds, could be extended to an animal with an auditory system more closely related to the human system and action potential VIIIth nerve data recorded in response to speech sounds. The results will suggest new sequences to code sounds for cochlear implants that could be tested in a clinical environment. In the future, these types of codes could enhance the effectiveness of implants in patients who are not significantly aided by the cochlear prosthesis using the present signal processing techniques.
