Thirty million Americans suffer from hearing loss. Several generations of the cochlear implant have been developed to help restore hearing to people with severe-to- profound hearing loss. The cochlear implant is surgically implanted in the inner ear and stimulates the auditory nerve with electric currents that encode certain acoustic features. We will review speech processing strategies that have been used for cochlear implants and discuss their performance and limitations. In general, the present cochlear implant users can understand 70-80% speech in quiet, allowing them to communicate over the telephone. However, their speech performance drops dramatically in a noisy environment. In addition, they cannot fully appreciate music. We will present a new sound processing strategy, which can dynamically encode the slowing varying as well as fast varying components in speech and music sounds.
Several experiments were performed to establish the utilities and limitations of the new speech strategy. A frequency modulation detection experiment revealed that cochlear implant users could only discern dynamic rate changes at low frequencies. Novel pulse trains, mimicking the stochastic neuron-firing pattern in normal acoustic hearing, were employed to improve rate pitch perception in cochlear implant users. Electrode ranking and pitch estimation experiments were conducted to examine the relative contributions of rate and place cues to pitch perception in electric hearing. Finally, the novel speech strategy was implemented on a real-time DSP processor, in which pitch information is encoded by varying the stimulation rate and the site of the stimulation. Results showed that optimally combining the rate and place pitch cues can improve the current cochlear-implant users melody recognition by as much as 36%. The new strategy has a great potential to improve cochlear implant users'