Effects of noise and noise reduction processing on the operation of the Nucleus-22 cochlear implant processor

Cochlear implants, like other types of auditory sensory aids, become increasingly ineffective with increasing ambient noise levels. One method of signal processing to reduce additive random wideband noise, the INTEL method, has been used to good effect as an input preprocessor for the Nucleus-22 cochlear implant. The implant's own signal processor estimates and encodes pitch frequency and the frequencies of Formants 1 and 2. The study reported here shows that additive noise results in substantial deviations in formant frequency estimates from those that would be observed in the absence of noise. When noisy speech is preprocessed by the INTEL method to reduce noise intensity, the deviations in the frequency estimates for Formant 2 are substantially reduced.