Cochlear Implant Atlas
CI Atlas · Speech-Coding Strategies: The Complete Lineage · Module 10

10ACE: Spectral Maxima Meet High Rate

The Advanced Combination Encoder fused SPEAK-style spectral-peak selection with CIS-style high-rate non-overlapping pulses, becoming the default strategy of the Nucleus system. ACE is the present-day commercial synthesis of the n-of-m lineage — and a useful baseline for understanding what comes after it.

TThe combination idea

ACE combines two principles: SPEAK-style spectral-peak (n-of-m) selection of the highest-energy bands and CIS-style high-rate non-overlapping interleaved stimulation ACE and fixed n-of-m designs are described as essentially identical, with n fixed across frames (unlike SPEAK's adaptive n) ACE selects 8-12 maxima from 10-12 FFT-based bands, or with a typically 22-band filter bank matching 22 physical electrodes ACE is the default / first-choice strategy for the Cochlear Nucleus system.[2006][2001]

+41.1Percentage-point ACE advantage on words in +10 dB SNR noise (pediatric) [2002]
10/11Adults preferring ACE at study end (Kiefer 2001) [2001]

CThe ACE signal chain

The chain is: mic to front-end to a band-pass filter bank to per-band envelope detection to per-frame selection of the n highest-amplitude bands (inactivating the rest) to compression and T/C-level mapping to high-rate interleaved biphasic pulses on the selected electrodes Pulses are delivered base-to-apex within each stimulation cycle, where the cycle is the inverse of the channel stimulation rate The default stimulation rate for Nucleus implants using ACE is about 900 pps On the Nucleus-24 the rate is selectable from 250 to 2400 pps per channel, up to a maximum of 14,400 pps across all channels.[2006][2001]

Pick n of 22, fire base→apex, repeat fast

base (high f)apex (low f)n = 10 fired · 9,000 pps aggregate

ACE — the Advanced Combination Encoder — is SPEAK's peak-picking run at CIS-like speed. Each cycle it selects the n highest of 22 bands (default 8–12) and stimulates them in turn, then immediately repeats. Raising n broadens spectral coverage; raising the per-channel rate samples each envelope more finely — and the two multiply into the aggregate rate the link must carry. ACE's default ~900 pps is what let it combine SPEAK's spectral selection with CIS's temporal fidelity. Schematic.

CThe Nucleus-24 platform

The Nucleus-24 has 22 intracochlear electrodes plus 2 extracochlear reference electrodes (a platinum ball under the temporalis and a plate on the receiver/stimulator) ACE has a higher rate and a larger peak-selection range than SPEAK The Nucleus-24 is also programmable with CIS using fixed amplitudes over 10-12 bands Recent comparisons generally show ACE superior to SPEAK and roughly comparable to CIS in challenging conditions.[2006][2001]

TWhy the lineage continues past ACE

ACE still uses raw-energy (amplitude) maxima selection, which may not match perceptual relevance ACE is envelope-based and does not encode temporal fine structure for pitch and music Selecting bands by raw energy motivated a psychoacoustic-masking selection rule (PACE/MP3000) Despite differing strategies, present commercial devices show no significant overall performance difference, underscoring that ACE is a strong but non-final synthesis.[2005][2004]

Pulse rate vs the same envelope

SPEAK ~250 ppsACE ~900 pps

SPEAK and ACE select spectral maxima the same way; they differ in rate. At SPEAK's ~250 pps only a handful of pulses fall under each envelope cycle, so fast amplitude changes are smoothed over. ACE's ~900 pps lays down far more pulses, capturing the envelope's shape — and with it more of the temporal detail that helps in noise and with music. Same selection rule, finer temporal grain. Schematic.

TDoes the high rate itself help, or just the n-of-m design?

ACE applies SPEAK-style n-of-m peak-picking at high per-channel rates; in the Nucleus 24 it is commonly run near 720-1800 pps/ch (Holden 2002). Despite the rate increase, controlled rate studies show no consistent group benefit of higher ACE rate: 720 vs 1800 pps/ch gave no conclusive advantage (Holden 2002), and 250/807 pps/ch were preferred over 1615 (Vandali 2000). Across implant types, speech recognition is largely flat with rate in quiet; only the lowest rates and non-interleaved high rates were significantly poorer (Friesen 2005). The only significant rate effect in one careful study was vowels in quiet, a small +5.23 pp rise from 600 to 4800 ppse, with consonants, words, and sentences unchanged (Shannon 2011). So ACE's clinical edge over SPEAK is attributed mainly to its faster temporal sampling and finer n-of-m updating rather than raw pulse rate per se; in children ACE beat SPEAK by +22.2 pp on words in quiet (Pasanisi 2002).[2002][2000]

TBy the numbers

ACE vs SPEAK: Pediatric Word and Phrase Recognition in Quiet and Noise

0255075100Percent correctWords, quietPhrases, quietWords, +10 dB SNRPhrases, +10 dB SNR
Material and conditionPhrases, +10 dB SNRSPEAK41.1%ACE82.2%

ACE keeps SPEAK's n-of-m peak-picking but runs it at far higher per-channel rates. In Nucleus CI24M children the ACE advantage over SPEAK was large and grew in noise: words +22.2 pp in quiet (68.8%->91%) but +41.1 pp at +10 dB SNR (43.3%->84.4%). In adults, 7/12 (Skinner 2002) and 10/11 (Kiefer 2001) preferred ACE. Verified means from Pasanisi 2002 (n=9 children).

Case 14.10 · ACE
A newly implanted Nucleus-24 adult is fitted with ACE at 900 pps selecting 10 of 22 bands. The clinician is asked how ACE differs from the SPEAK strategy the clinic used a decade earlier.

What is the most accurate description of the ACE-versus-SPEAK difference?

Self-assessment — Module 102 questions
Question 1

What is the default stimulation rate for Nucleus cochlear implants using ACE?

Question 2

How many electrodes does the Nucleus-24 have, and of what type?

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