Wk5a - Other Speech Processing Strategies Flashcards
Name two variations of CIS
SPEAK (Spectral Peak; obsolete)
ACE/n-of-m
What does SPEAK stand for? How does it work?
Spectral Peak
- it stimulates a variable number of electrodes at a low pulse rate (~250 pps) compared to CIS (~1000 pps)
- the electrodes corresponding to the 6-10 highest amplitude channels are stimulated
With the SPEAK speech processing strategy, the number of channels is adaptive. What 2 things determine the number of channels used?
Input level
Spectrum
What speech processing strategy superseded SPEAK?
ACE
Describe how the SPEAK processing strategy processes an acoustic signal?
- The waveform is passed through a 22 channel bandpass filter bank
- At a single moment in time, the filter-bank output envelopes are evaluated and the 6 channels with the largest output results in stimulation of the corresponding electrodes
- On another cycle, other channels might be selected
What does ACE stand for?
Advanced Combination Encoder
How does ACE differ from SPEAK?
- combines high pulse rate (like CIS) with channel selection
- only stimulates the electrodes corresponding to the highest n filter bank channel levels (e.g. 8 out of 16) on each cycle
- the number of active electrodes is NON-ADAPTIVE (can be changed in the CI program by the audiologist, not changing with each cycle)
What is the name of the strategy of which ACE is a proprietary name of?
n-of-m
Describe the n-of-m strategy?
- on each cycle, the n largest filter bank outputs will results in the stimulation of the corresponding n electrodes, out of the total m electrodes of the implant
- n is a variable programming parameter, but does not vary c/ the incoming signal (non-adaptive)
What are 2 benefits of n-of-m?
- maintaining fine structure
- noise reduction?
CIS mostly eliminates fine structure info, focussing mainly on the _______
Envelope
What 3 strategies did Med-EL use to attempt to maintain fine structure information?
- FSP (fine structure processing)
- FS4
- FS4-p
How did Med-EL’s fine structure processing strategies work?
- high frequency channels use CIS variants CIS+ or HDCIS (overlapping band-pass filters and Hilbert envelope extractor)
- the fine structure was maintained in the 2-4 low frequency channels, where pulses were only emitted when the waveform crossed zero in a positive direction
- supposedly gives listener access to additional low-freq fine structure info, helpful for pitch perception and SIN
With Med-EL’s FSP strategy, pulse trains were generated at each ascending zero crossing of the acoustic signal. What might determine which electrodes were programmed to receive this signal?
Depth of insertion - we have a rate-limit for AM sensitive of 250-300 hz
- deeper insertion = more fine structure electrodes can be employed (up to 4 c/ FSP)
FS4 was programmed to deliver fine structure stimulation to the 4 most ____ electrodes up to ____ Hz
Apical; 1000 Hz
Between HDCIS and FS4, which demonstrates more fine structure in the first 4 electrodes/channels?
FS4
Which of Med-El’s 4 strategies used sharp filter banks and a bandwidth b/w 250-8500 Hz?
CIS+
What was CIS+ upgraded to?
HDCIS - the same as CIS+, but with a bell-shaped filter bank
Which Med-El strategy included fine structure in the 2-4 apical electrodes up to 470 Hz and extended bandwidth down to 100 Hz
FSP
How is FS4 different from FSP?
- 4 apical electrodes always stimulated to provide fine structure (vs 2-4)
- fine structure bandwidth increased to up to 1000 Hz
How is FS4p different from FS4?
- it’s FS4 with parallel stimulation
How does a sharp filter bank differ from overlapping?
- sharp filter banks have small amount of info being carried by neighbouring electrodes (sharply tuned)
- bell-shaped will overlap info b/w diff channels (e.g. 1kHz info carried mostly by one filter, but also a little by each on either side
- trend is to overlapping filters (so sound “fades” b/w electrodes, instead of “jumping”)
What is current steering?
Activation of multiple electrodes simultaneously
- can stimulate diff (but maybe overlapping) populations of neurons than those from activating any single electrode
- on-purpose channel interaction
What changes can we expect to see to the typical bell-shaped electric field by simultaneously stimulating adjacent electrodes with the same bipolar stimulus with half the current/amplitude?
We would see a wider, flatter stimulus
(whereas if one electrode were stimulated with the same current and electrode 2 with half amplitude and reversed polarity, we would see the same bell shape, but with sharper/ more precise stimulation)
How is “on purpose” interaction (current steering) being used?
- to generate “virtual” channels b/w electrodes
- using pairs of electrodes, they can vary the ratio of simultaneous pulse amplitudes for each virtual channel
E.g. Advanced Bionics’ HiRes120 (120 channels using only 16 real electrodes)
E.g. Using current steering to create “virtual” channels, describe the current inputs for the following:
A virtual channel is centred directly over electrode 1. How much current is it receiving from electrode 2?
0%; 100% of the input is from electrode 1
E.g. Using current steering to create “virtual” channels, describe the current inputs for the following:
A virtual channel is centred a quarter of the way between electrodes 1 and 2. How much current is it receiving from electrode 2?
25%; 75% of the current is from electrode 1.
New speech strategies may only work with new generation implants, or may require ________ or ________ (multiple words)
- firmware updates
- new external hardware
Acoustic pre-processing algorithms from hearing aids can generally be used in CI’s ____ (prior to, after) speech coding
Prior to
