Signal analysis by using FIR filter banks in cochlear implant prostheses

The use of auditory models in cochlear implant speech processing strategies aims to improve cochlear implantee's perception of speech. Current speech processing strategies for cochlear implants use an IIR filter bank which decomposes the audio signals into multiple frequency bands each associated with one electrode. This paper describes a 16-channel non-uniformly spaced, high stop band attenuation (≅ −60dB) finite impulse response (FIR) filter bank model for cochlear implants and how speech intelligibility is evaluated as a function of various parameters in cochlear implant processors. The design has the advantages of higher stop band attenuation and linear phase frequency response. The filter bank implemented is successfully tested using speech signals extracted from TIMIT database and their band energies are calculated which are useful in estimating the stimulation pulse level to be generated and presented to the cochlear electrodes.