We describe our first-hand experience with a cochlear implant (CI), being both a recent recipient and a hearing researcher. We note the promising loudness, but very unpleasant distortion, which makes understanding speech difficult in many environments, including in noise, on the phone or through the radio. We also discuss the extreme unpleasantness of music, which makes recognizing familiar melodies very difficult. We investigate the causes of the above problems through mathematical analysis and computer simulations of sound mixtures, and find that surprisingly, the culprit appears to be non-biological in origin, but primarily due to the envelope-based signal processing algorithms currently used. This distortion is generated before the signal even enters the cochlea. Hence, the long-held belief that inter-electrode interference or current spreading is the cause, appears incorrect. We explain that envelope processing may have been originally instituted based on an inaccurate understanding of the role of place coding vs. temporal coding, or alternatively, because of an incorrect analogy to radio modulation theory. On the basis of our analysis, we suggest immediate concrete steps, some possibly in firmware alone, that may lead to a much improved experience.
A Computational Model of Cellular Mechanisms of Temporal Coding in the Medial Geniculate Body (MGB)
