Cochlear Implant Atlas
CI Atlas · Speech-Coding Strategies: The Complete Lineage · Module 07

7Multipeak (MPEAK): Formants Plus High-Frequency Bands

Multipeak was the high-water mark of the feature-extraction era. It kept F0/F1/F2 formant tracking but bolted on three fixed high-frequency band-pass filters to recover consonant cues. It improved consonant identification — yet its formant-extraction errors in noise foretold the end of the feature-extraction philosophy.

TThe MPEAK design

MPEAK refines F0/F1/F2 by additionally extracting high-frequency information through three extra fixed band-pass filters to improve consonant perception It combines formant tracking by zero-crossing detection with high-frequency envelope detection The three added high-frequency bands are 2000-2800 Hz, 2800-4000 Hz and 4000-6000 Hz MPEAK was implemented on the Nucleus cochlear implant from Cochlear Corporation.[2008][1993]

67.5%MPEAK consonant identification (n=9) [1999]
8.7Percentage-point SPEAK consonant advantage over MPEAK (p<0.001) [1999]

CThe full MPEAK chain

The chain is: mic to AGC, then F0 via a 270 Hz LPF, F1/A1 via a 300-1000 Hz band, F2/A2 via an 800-4000 Hz band, plus three high-frequency envelope-detector bands at 2000-2800, 2800-4000 and 4000-6000 Hz MPEAK stimulates four electrodes; voiced segments use F0 pps and unvoiced segments use about 250 pps quasi-random For voiced sounds it stimulates the F1 and F2 electrodes plus high-frequency electrodes 4 (2800-4000 Hz) and 7 (2000-2800 Hz); the 4-6 kHz electrode is not stimulated because there is little energy above 4 kHz For unvoiced sounds it stimulates high-frequency electrodes 1 (4000-6000 Hz), 4 and 7 plus the F2 electrode; the F1 electrode is not stimulated because there is little energy below 1 kHz.[1993][2008]

MPEAK fires four electrodes — which four depends on voicing

F1F1 formantF2F2 formantE72.0–2.8 kHzE42.8–4.0 kHzE14.0–6.0 kHzapexbase

Multipeak kept the F0/F1/F2 formant trackers but added three fixed high-frequency bands (≈2–2.8, 2.8–4, 4–6 kHz) to carry consonant energy. Four electrodes fire each cycle, and the rule shifts with voicing: for voiced sounds it favours F1, F2 and the lower high bands; for unvoiced sounds, where energy sits high and F1 is weak, it drops F1 and recruits the top band. Better consonants — but still hostage to correct formant estimation. Schematic.

TWhat MPEAK achieved

MPEAK improved consonant identification relative to the pure formant strategies It used voicing-dependent electrode selection, allocating high-frequency electrodes differently for voiced versus unvoiced segments The three dedicated high-frequency bands directly targeted the consonant information missing from F0/F1/F2 F2 was extracted with an 800-4000 Hz band-pass filter.[2008][1993]

CWhy MPEAK was superseded

MPEAK tends to make formant-extraction errors when speech is embedded in noise, degrading performance This noise vulnerability is the generic failure mode of feature-extraction algorithms that rely on estimating specific parameters The limitation motivated the spectral-maxima approach (SMSP), which extracts no explicit features and instead transmits the largest spectral peaks The shift from MPEAK to spectral-maxima marks the end of the explicit feature-extraction era.[1993][2006]

Add noise — watch the F2 estimate break up

true F2estimatetracking error ≈ 1%

A feature-extraction processor must estimate the formants before it can code them — and estimation is exactly what noise destroys. As the background rises, the F2 tracker locks onto noise peaks and jumps erratically (red), sending the wrong electrode and smearing the percept. Because the discarded spectrum is gone, there is nothing to fall back on. This single failure mode is why the field abandoned formant tracking for strategies that pick spectral maxima directly. Schematic.

TBy the numbers

MULTIPEAK vs SPEAK: Vowel and Consonant Identification

020406080Percent correctVowelsConsonants
Phoneme classConsonantsMPEAK67.5%SPEAK76.2%

MULTIPEAK (MPEAK) added fixed high-frequency consonant bands on top of formant tracking. Head-to-head with the later spectral-maxima SPEAK strategy, MPEAK matched it on vowels (72.3% vs 73.4%) but trailed on consonants (67.5% vs 76.2%, p<0.001) because SPEAK transmitted more place/F2-transition information. Verified means from Skinner/Fourakis (n=9).

Case 14.7 · Multipeak (MPEAK)
Two recipients are compared in quiet and in noise. On MPEAK both do well in quiet, but in cafeteria noise one drops sharply while a third recipient on a spectral-maxima strategy holds up better.

What is the most likely reason MPEAK degrades more in noise?

Self-assessment — Module 72 questions
Question 1

What did MPEAK add to the F0/F1/F2 strategy?

Question 2

What weakness of MPEAK most directly motivated the spectral-maxima approach?

Tracked locally in your browser — see /progress for the dashboard.