4Filter banks & the place code
The cochlea's first trick is to be a frequency analyser: it lays sound out by pitch along its length, low frequencies at the apex and high at the base. The implant copies this with a bank of electronic filters — one per channel — and a simple rule: route each frequency band to the electrode sitting at the matching place on the array. That mapping is the implant's primary way of conveying pitch, and getting it right matters, because if the band sent to an electrode does not match the cochlear place that electrode actually stimulates, the brain receives a distorted, transposed version of the spectrum. This module covers how the filter bank rebuilds the place code, and what happens when the map and the anatomy disagree.
TThe cochlea as a frequency analyser
As the physiology chapter showed (Chapter 2), the healthy cochlea performs a running frequency analysis, with each place along the basilar membrane tuned to a characteristic frequency — the tonotopic or place code. The implant cannot move the basilar membrane, but it can imitate the result: assign frequencies to cochlear places electrically.
CThe filter bank
The processor passes the incoming sound through a bank of band-pass filters, one for each channel, dividing the spectrum into contiguous frequency bands (often allocated to follow the cochlea's roughly logarithmic Greenwood map). The output of each filter drives one electrode. A low band drives an apical electrode, a high band a basal one — and complex sounds like speech light up many channels at once, in a moving pattern.
CThe place code restored
This place-by-place mapping is the implant's primary pitch cue: higher-pitched sounds excite more basal electrodes, lower ones more apical. It is coarse — only a handful of distinguishable places — and it lacks the fine temporal cue that sharpens normal pitch, which is why implant pitch perception is limited. But it is enough to carry the spectral shape of speech, the formant pattern that distinguishes one vowel from another.
CWhen the map is wrong
The mapping assumes the electrode sits at the cochlear place its assigned frequency belongs to — but a real array may be inserted only part-way, or sit in a cochlea of unusual length, so an electrode delivering “1 kHz” actually lies at a place tuned to a different frequency. This frequency-to-place mismatch transposes the spectrum, and the brain must acclimatise to it (a plasticity task, Chapter 3). Image-guided and anatomy-based fitting aim to reduce the mismatch by tailoring the allocation to where the electrodes really are.
What explains the percept, and what helps over time?
How does the implant recreate the cochlea's place code?
What is a frequency-to-place mismatch and why does it matter?