ABSTRACT
This chapter suggests that implant speech performance and simulations of implant speech processors are valuable tools for understanding normal speech pattern recognition. The performance of patients with cochlear implants has posed an interesting problem for theories of speech recognition. Patients can recognize running speech well enough to converse on the telephone, even with only four electrodes, spaced several millimeters apart in the scala tympani, each stimulating a broad region of auditory neurons in a highly unnatural manner. In normal acoustic hearing, speech is processed by the auditory periphery, which delivers a complex pattern of neural information to the brain. Central pattern recognition systems are trained to recognize and distinguish these patterns over a lifetime of speech communication. Speech was divided into several spectral bands by bandpass filtering. The speech envelope was extracted from each band by full-wave rectification and low-pass filtering.
