n-of-m and ACE (peak-picking) strategies

n-of-m strategies (SPEAK, and its fixed-rate descendant ACE, Advanced Combination Encoders) add a channel-selection step to the CIS idea: of the m analysis bands, only the n with the largest instantaneous envelopes are stimulated each cycle (for example 8-of-22). This concentrates the limited stimulation budget on spectral peaks (formants), reduces channel interaction by not driving low-energy bands, and permits many analysis filters (finer place resolution) without a prohibitive total pulse rate. ACE is the default strategy in Cochlear-brand devices and is, empirically, comparable to CIS and HiRes for speech.

The tradeoff is that peak-picking is a lossy, nonlinear, level-dependent spectral selection: it can discard low-level spectral detail, distort loudness relations, and its benefit over plain CIS is modest and listener-dependent. n-of-m is best seen as a pragmatic compromise for a device whose true bottleneck is the small number of effectively independent channels, not the selection rule.